Image and Graphics Technologies and Applications

14th Conference on Image and Graphics

Technologies and Applications IGTA 2019 Beijing
China April 19 20 2019 Revised Selected Papers
Yongtian Wang download
https://textbookfull.com/product/image-and-graphics-technologies-
and-applications-14th-conference-on-image-and-graphics-
technologies-and-applications-igta-2019-beijing-china-
april-19-20-2019-revised-selected-papers-yongtian-wang/

Download more ebook from https://textbookfull.com

We believe these products will be a great fit for you. Click
the link to download now, or visit textbookfull.com
to discover even more!

Advances in Image and Graphics Technologies 12th

Chinese conference IGTA 2017 Beijing China June 30 July
1 2017 Revised Selected Papers 1st Edition Yongtian
Wang
https://textbookfull.com/product/advances-in-image-and-graphics-
technologies-12th-chinese-conference-igta-2017-beijing-china-
june-30-july-1-2017-revised-selected-papers-1st-edition-yongtian-
wang/

Advances in Image and Graphics Technologies 10th

Chinese Conference IGTA 2015 Beijing China June 19 20
2015 Proceedings 1st Edition Tieniu Tan

https://textbookfull.com/product/advances-in-image-and-graphics-
technologies-10th-chinese-conference-igta-2015-beijing-china-
june-19-20-2015-proceedings-1st-edition-tieniu-tan/

Advances in Image and Graphics Technologies 11th

Chinese Conference IGTA 2016 Beijing China July 8 9
2016 Proceedings Tieniu Tan

https://textbookfull.com/product/advances-in-image-and-graphics-
technologies-11th-chinese-conference-igta-2016-beijing-china-
july-8-9-2016-proceedings-tieniu-tan/

Image and Graphics 10th International Conference ICIG

2019 Beijing China August 23 25 2019 Proceedings Part
III Yao Zhao

https://textbookfull.com/product/image-and-graphics-10th-
international-conference-icig-2019-beijing-china-
august-23-25-2019-proceedings-part-iii-yao-zhao/
Image and Graphics 9th International Conference ICIG
2017 Shanghai China September 13 15 2017 Revised
Selected Papers Part III 1st Edition Yao Zhao

https://textbookfull.com/product/image-and-graphics-9th-
international-conference-icig-2017-shanghai-china-
september-13-15-2017-revised-selected-papers-part-iii-1st-
edition-yao-zhao/

Simulation and Modeling Methodologies, Technologies and

Applications: 9th International Conference, SIMULTECH
2019 Prague, Czech Republic, July 29-31, 2019, Revised
Selected Papers Mohammad S. Obaidat
https://textbookfull.com/product/simulation-and-modeling-
methodologies-technologies-and-applications-9th-international-
conference-simultech-2019-prague-czech-republic-
july-29-31-2019-revised-selected-papers-mohammad-s-obaidat/

Software Technologies 14th International Conference

ICSOFT 2019 Prague Czech Republic July 26 28 2019
Revised Selected Papers Marten Van Sinderen

https://textbookfull.com/product/software-technologies-14th-
international-conference-icsoft-2019-prague-czech-republic-
july-26-28-2019-revised-selected-papers-marten-van-sinderen/

Advances in Smart Technologies Applications and Case

Studies Selected Papers from the First International
Conference on Smart Information and Communication
Technologies SmartICT 2019 September 26 28 2019 Saidia
Morocco Ali El Moussati
https://textbookfull.com/product/advances-in-smart-technologies-
applications-and-case-studies-selected-papers-from-the-first-
international-conference-on-smart-information-and-communication-
technologies-smartict-2019-september-26-28-2/

Computer Supported Cooperative Work and Social

Computing 14th CCF Conference ChineseCSCW 2019 Kunming
China August 16 18 2019 Revised Selected Papers Yuqing
Sun
https://textbookfull.com/product/computer-supported-cooperative-
work-and-social-computing-14th-ccf-conference-
chinesecscw-2019-kunming-china-august-16-18-2019-revised-
Yongtian Wang
Qingmin Huang
Yuxin Peng (Eds.)

Communications in Computer and Information Science 1043

Image and Graphics

Technologies
and Applications
14th Conference on Image and Graphics
Technologies and Applications, IGTA 2019
Beijing, China, April 19–20, 2019
Revised Selected Papers
Communications
in Computer and Information Science 1043
Commenced Publication in 2007
Founding and Former Series Editors:
Phoebe Chen, Alfredo Cuzzocrea, Xiaoyong Du, Orhun Kara, Ting Liu,
Krishna M. Sivalingam, Dominik Ślęzak, Takashi Washio, and Xiaokang Yang

Editorial Board Members

Simone Diniz Junqueira Barbosa
Pontifical Catholic University of Rio de Janeiro (PUC-Rio),
Rio de Janeiro, Brazil
Joaquim Filipe
Polytechnic Institute of Setúbal, Setúbal, Portugal
Ashish Ghosh
Indian Statistical Institute, Kolkata, India
Igor Kotenko
St. Petersburg Institute for Informatics and Automation of the Russian
Academy of Sciences, St. Petersburg, Russia
Junsong Yuan
University at Buffalo, The State University of New York, Buffalo, NY, USA
Lizhu Zhou
Tsinghua University, Beijing, China
More information about this series at http://www.springer.com/series/7899
Yongtian Wang Qingmin Huang
• •

Yuxin Peng (Eds.)

Image and Graphics

Technologies
and Applications
14th Conference on Image and Graphics
Technologies and Applications, IGTA 2019
Beijing, China, April 19–20, 2019
Revised Selected Papers

123
Editors
Yongtian Wang Qingmin Huang
Beijing Institute of Technology University of Chinese Academy of Science
Beijing, China Beijing, China
Yuxin Peng
Institute of Computer Science
and Technology
Peking University
Beijing, China

ISSN 1865-0929 ISSN 1865-0937 (electronic)

Communications in Computer and Information Science
ISBN 978-981-13-9916-9 ISBN 978-981-13-9917-6 (eBook)
https://doi.org/10.1007/978-981-13-9917-6
© Springer Nature Singapore Pte Ltd. 2019
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, expressed or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
 Preface

It was a pleasure for us to organize the 14th Conference on Image and Graphics
Technologies and Applications (IGTA 2019). The conference was organized under
the auspices of the Beijing Society of Image and Graphics, and was held during
April 19–20, 2019, at the University of Chinese Academy of Sciences in Beijing,
China.
IGTA is a professional conference and a premier forum for image processing,
computer graphics and related topics, including but not limited to image analysis and
understanding, computer vision and pattern recognition, big data mining, virtual reality
and augmented reality, as well as image technology applications.
IGTA 2019 collected over 152 submissions for technical presentation from different
countries and regions of the world. Each of the manuscripts was assessed by at least
two reviewers, with some of them being assessed by three reviewers. After careful
evaluation, a total of 66 manuscripts were selected for oral and poster presentations.
The conference program included keynote presentations, oral papers, workshops,
posters, demos, and exhibitions. The most recent progress in the field of images and
graphics was reported. I firmly believe that the papers included in the IGTA 2019
proceedings will provide valuable reference information to scientists and engineers in
the relevant fields.
On behalf of conference general chairs, I would like to heartily thank our supporters
and committee members for all they have done for this conference. Thanks go to all
authors for their contributions, especially to those who traveled great distances and
took time from their busy schedules to attend the conference. I am also grateful to the
Springer staff for their support and collaboration in publishing these proceedings.
Thanks also go to the reviewers who completed the reviewing process on time. Finally,
I would like to express our gratitude to our host, professors and students from the
University of Chinese Academy of Sciences, who took care of the local arrangements
for the conference, and to all the attendees.

April 2019 Yongtian Wang

 Organization

General Conference Chairs

Yongtian Wang Beijing Institute of Technology, China
Qingming Huang University of Chinese Academy of Sciences, China

Executive and Coordination Committee

Guoping Wang Peking University, China
Chaowu Chen The First Research Institute of the Ministry of Public
Security of P.R.C., China
Mingquan Zhou Beijing Normal University, China
Shengjin Wang Tsinghua University, China
Chenglin Liu Institute of Automation, Chinese Academy of Sciences,
China
Yao Zhao Beijing Jiaotong University, China
Zhiguo Jiang Beihang University, China

Program Committee Chairs

Yuxin Peng Peking University, China
Ran He Institute of Automation, Chinese Academy of Sciences,
China

Organizing Chairs
Yue Liu Beijing Institute of Technology, China
Weiqiang Liu University of Chinese Academy of Sciences, China
Xiaoru Yuan Peking University, China

Research Committee Chairs

Xiaohui Liang Beihang University, China
Xiangyang Ji Tsinghua University, China
Jian Yang Beijing Institute of Technology, China

Publicity and Exhibition Committee Chairs

Lei Yang Communication University of China, China
Fengjun Zhang Software Institute of the Chinese Academy of Sciences,
China
viii Organization

Program Committee
David Rempel University of California, USA
Zhao Chaoyi China National Institute of Standardization (CNIS), China
Stephen Bao University of Washington, USA
Jay Kapellusch University of Wisconsin, Milwaukee, USA
Lin Wanhong China Astronaut Research and Training Center, China
Henry Been-Lirn Duh La Trobe University, Australia
Takafumi Taketomi NAIST, Japan
Jeremy M. Wolfe Harvard Medical School, USA
Yiping Huang Taiwan University, China
Youngho Lee Mokpo National University, South Korea
Nobuchika Sakata Osaka University, Japan
Seokhee Jeon Kyunghee University, South Korea
Xiaoru Yuan Peking University, China
Ran He Institute of Automation, Chinese Academy of Sciences,
China
Jian Yang Beijing Institute of Technology, China
Xiangyang Ji Tsinghua University, China
Yue Liu Beijing Institute of Technology, China
Huimin Ma Tsinghua University, China
Liang Wang Institute of Automation, Chinese Academy of Sciences,
China
Huijie Zhao Beijing University of Aeronautics and Astronautics, China
Danpei Zhao Beijing University of Aeronautics and Astronautics, China
Cheng Yang Communication University of China, China
Jun Yan Journal of Image and Graphics
Shihong Xia Institute of Computing Technology, Chinese Academy
of Sciences, China
Weiqun Cao Beijing Forestry University, China
Kaichang Di Institute of Remote Sensing and Digital Earth,
Chinese Academy of Sciences, China
Xucheng Yin University of Science and Technology Beijing, China
Fuping Gan Ministry of Land and Resources of the People’s Republic
of China
Xueqiang Lv Beijing Information Science and Technology University,
China
Jianbo Liu Communication University of China, China
HuaLin Tsinghua University, China
Xiaozhu Lin Beijing Institute of Petrochemical Technology, China
Hua Li Institute of Computing Technology, Chinese Academy
of Sciences, China
Jing Dong Institute of Automation, Chinese Academy of Sciences,
China
Yankui Sun Tsinghua University, China
Li Zhuo Beijing University of Technology, China
 Organization ix

Qingyuan Li Chinese Academy of Surveying and Mapping, China

JiazhengYuan Beijing Union University, China
Yiding Wang North China University of Technology, China
Aiwu Zhang Capital Normal University, China
Mingzhi Cheng Beijing Institute of Graphic Communication, China
Yahui Wang Beijing University of Civil Engineering and Architecture,
China
Siwei Ma Peking University, China
Liang Liu Beijing University of Posts and Telecommunications,
China
Bin Liao North China Electric Power University, China
 Contents

Human Fungal Infection Image Classification Based on Convolutional

Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Yuan Zhou, Yanxia Feng, and Haiying Zhang

Improved RPN for Single Targets Detection Based on the Anchor

Mask Net . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Mingjie Li, Youqian Feng, Zhonghai Yin, Cheng Zhou,
and Fanghao Dong

Design and Analysis of Object Behavior in Media Content-User

Relationship Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Shan Liu and Kun Huang

A Method of Penicillin Bottle Defect Inspection Based on BP

Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Yangbo Feng, Tinglong Tang, and Shengyong Chen

An End-to-End Pyramid Convolutional Neural Network for Dehazing . . . . . . 41

Chuandong Yang, Zhen Liu, Songnan Liu, Jie Qin, and Dong Chen

Dynamic Community Detection and Evolution Analysis. . . . . . . . . . . . . . . . 51

Shan Liu and Rui Tang

Object Detection and Segmentation Method for Multi-category Armored

Targets Based on CNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Dong Chen, Chuandong Yang, Zhen Liu, Xiaolong Zhang,
and Shengbin Shi

Inter-frame Relationship Graph Based Near-Duplicate Video Clip

Detection Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Xinbo Ai, Yinan He, Yanzhu Hu, and Wenjia Tian

The High Speed 3D Measurement Based on Interval Line Structured Light

Method for Translucent Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Huijie Zhao, Xiaochun Diao, Hongzhi Jiang, and Xudong Li

Pixel and Channel Attention Network for Person Re-identification . . . . . . . . 91

Minjie Wang, Xian Li, Jiahuan Zhang, Haoyu Zhou, Lei Lei,
and Banghua Yang

Infrared and Visible Image Matching Algorithm Based on SIFT and LDB . . . 105
Lirui Zhang, Min Dai, and Jinwen Tian
xii Contents

The Overview of 2D to 3D Automatic Conversion . . . . . . . . . . . . . . . . . . . 116

Yunlong Cheng, Yun Dong, and Jiawei Tan

Sugarcane Node Identification Based on Structured Learning Model . . . . . . . 125

Xiao Hu, Shuqin Li, Meili Wang, Changyou Shi, Jing Shang,
Yunhua Pei, Ganran Deng, and Deqiang Zhou

3D Human Pose Estimation with Grouping Regression . . . . . . . . . . . . . . . . 138

Xuesheng He, Huabin Wang, Yuan Qin, and Liang Tao

Vascular Model Editing for 3D Printing Based on Implicit Functions . . . . . . 150

Beizhan Wang, Qichao Ge, Qingqi Hong, Yangjing Li, Kunhong Liu,
and Ziyou Jiang

Design and Implementation of Hardware Accelerator for Gaussian Filter

Based on HLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Zonghao Tian, Zhen Liu, Shuguang Wang, and Dong Chen

Fusion of Global and Local Gaussian-Hermite Moments

for Face Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Guojie Song, Dan He, Puchun Chen, Jidong Tian, Bin Zhou, and Li Luo

A Visual-Inertial Information Fusion Method for SLAM

Front-End Odometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
Xinguo Yu, Tai Li, and Zhizhong Zeng

A Multiscale Image Denoising Algorithm Based on Dilated Residual

Convolution Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Chang Liu, Zhaowei Shang, and Anyong Qin

Model-Driven Dynamic Visualization of Spatiotemporal Data in GIS . . . . . . 204

Weiyi Kong, Li Yang, Jianlong Ren, Chun Zuo, and Fengjun Zhang

DCNN Transfer Learning and Multi-model Integration for Disease

and Weed Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
Jingxian Wang, Miao Li, Jian Zhang, WeiHui Zeng,
and XuanJiang Yang

3D Human Pose Lifting: From Joint Position to Joint Rotation . . . . . . . . . . . 228

Zeye Wu and Wujun Che

Estimation of 6Dof Pose Using Image Mask and Bounding Box . . . . . . . . . . 238
Yibo Cui, Pengyuan Liu, and Junning Zhang

Multi-Attention Network for 2D Face Alignment in the Wild . . . . . . . . . . . . 246

Xin Liu, Huabin Wang, Rui Cheng, Xiang Yan, and Liang Tao

Virtual-Real Fusion Processing Based on Parallel Distribution . . . . . . . . . . . 256

Hongyu Zhai, Tongtong Zhang, and Richeng Xu
 Contents xiii

Exploring a Facial Defect Skin Detection Algorithm with Probability

Distribution Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
Xingyuan Ren, Xiangyuan Qi, and Zhen Wang

Universal Framework of Seals Erasing with Generative

Adversarial Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
ZiQiang Chen, ZhenYu Ding, and ShiQing Wang

Pupil Segmentation Method Based on Treasure Theory

and Region Growing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
Kang Yao and Weiwei Fu

Object Recognition and Localization Base on Binocular Vision . . . . . . . . . . 300

Rongjie Duan, Shizhong Li, Zehui Yuan, and Ya Zhang

Effects of Dynamic Disparity on Visual Fatigue Caused by Watching

2D Videos in HMDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
Ruiying Shen, Dongdong Weng, Jie Guo, Hui Fang, and Haiyan Jiang

Detection of Small Moving Targets in Videos Using Skew Normal

Mixture Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
Yuhui Shao and Fang Dai

Spectral Pooling Based CNN Model Compression for Light Field

Depth-Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
Xuanyu Zhang, Guosheng Yang, and Jing Jiang

3D Surface Splicing Based on Principal Component Feature Extraction . . . . . 343

Kaiyue Li, Qingyu Meng, Mingquan Zhou, and Pengbo Zhou

An Automatic White Balance Algorithm Based on Pixel Luminance

and Chromaticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
Heding Xu, Hong Zhang, Bo Rao, Yifan Yang, and Zeyu Zhang

Attention-Based GAN for Single Image Super-Resolution . . . . . . . . . . . . . . 360

Dongqi Huo, Rong Wang, and Jianwei Ding

Tiny Vehicle Detection from UAV Imagery . . . . . . . . . . . . . . . . . . . . . . . . 370

Jinze Li, Rong Wang, and Jianwei Ding

SGM-Based Disparity Estimation Under Radiometric Variations . . . . . . . . . . 382

WeiMin Yuan, XiaoYan Tong, and Bin Xiao

A Survey of Chinese Character Style Transfer . . . . . . . . . . . . . . . . . . . . . . 392

Yifei Ma, Yuntai Dong, Kang Li, Jian Ren, Guohua Geng,
and Mingquan Zhou
xiv Contents

Deep Multi-resolution Feature Fusion Network

Based 3D Object Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
Yang Xiao, Yanxin Ma, and Jun Zhang

Construction of Game-Based Teaching Environment for Animation Scene

Scheduling Based on Virtual Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412
Xiaodong Wei and Yali Wu

An Improved Dark Object Subtraction Method for Atmospheric Correction

of Remote Sensing Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
Yu Wang, Xiaoyong Wang, Hongyan He, and Guoliang Tian

The Influence of Mobile Augmented Reality-Based Sandbox Games

on Chinese Characters Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
Xiaodong Wei, Guodong Yang, and Dongdong Weng

Impact of Teaching Aids Based on Virtual Role on Student’s Design

Thinking for Animation Course . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
Xiaodong Wei, Yang Liu, and Dongdong Weng

Classification of Imagined Digits via Brain-Computer Interface Based

on Electroencephalogram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
Melvin Harsono, Lie-quan Liang, Xin-wei Zheng, Forrest Fabian Jesse,
Yi-gang Cen, and Wen Jin

Plant Identification Based on Multi-branch Convolutional Neural Network

with Attention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472
Pengxi Li, Xiaoqing Gong, Xu Hu, Lianqi Shi, Xiaoting Xue, Jun Guo,
Pengfei Xu, and Daguang Gan

Verification Learning for Robust Visual Object Tracking . . . . . . . . . . . . . . . 482

Linyu Fei, Rong Wang, and Jianwei Ding

Convolutional Randomized Binary Features for Keypoints Recognition . . . . . 492

Jinming Zhang, Zuren Feng, and Gang Li

A Novel Multi-image Encryption Algorithm Based on the RDFrMT

and Cascaded Phase Retrieval. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
Haotian Yang, Tong Li, Sijiang Huang, Fei Qi, and Xuejing Kang

Analysis of Influence of the Spectral Channels of AWG on the Imaging

Quality of Planar Interferometric Imaging System . . . . . . . . . . . . . . . . . . . . 516
Gongmin Yu, Libing Jin, Feng Zhou, and Xiliang Tong

Remote Sensing Image Change Detection Algorithm Based

on BM3D and PCANet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524
Yi Wan, Yongjin Liu, Qunnie Peng, Feiran Jie, and Delie Ming
 Contents xv

Long-Tailed Contrastive Loss for Video-Based Person Re-identification. . . . . 532

Liqiang Bao

A Novel Image Super-Resolution Method Based on Cross Classification

Trees and Cascaded Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 541
Yuan Lv, Ziqi Wang, Peiqi Duan, and Xuejing Kang

Diagnostic Classification of Pulmonary Nodules Using a Multi-scale

and Multi-input DenseNet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553
Mengsong Wang, Hansheng Li, Yifan Wu, Qirong Bu, and Jun Feng

SAR Ship Detection Under Complex Background Based

on Attention Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
Chen Chen, Changhua Hu, Chuan He, Hong Pei, Zhenan Pang,
and Tong Zhao

A Beam Adjustment Algorithm Based on Star-Sensor Geometric Imaging

Link Model with Distance Constraint. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579
Guoliang Tian, Qiaolin Huang, Hongyan He, Yu Wang,
and Zhongqiu Xia

Dynamic Load Balancing Algorithm Based on Per-pixel Rendering

Cost Estimation for Parallel Ray Tracing on PC Clusters . . . . . . . . . . . . . . . 591
Chaozhi Yang, Chunyi Chen, Xiaojuan Hu, and Huamin Yang

Research on Warehouse Object Detection Algorithm Based on Fused

DenseNet and SSD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 602
Liangjie Chen, Fei Wang, Li Wang, and Lin Wang

Study on 3D Modeling of Complex Coal Mine Interface . . . . . . . . . . . . . . . 612

Qian-lin Dong, Yang-ming Jiang, Zhi-chao Hao, Wei-dong Li, Ke Wang,
Xing-dong Wang, and Qing-yuan Li

Enhanced Subtraction Image Guided Convolutional Neural Network

for Coronary Artery Segmentation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625
Jingfan Fan, Chenbin Du, Shuang Song, Weijian Cong, Aimin Hao,
and Jian Yang

Visual Tracking Based on Multi-cue Proposals and Long Short-Term

Features Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
Jiaming Wei, Huimin Ma, Ruiqi Lu, and Xiong Luo

MMRPet: Modular Mixed Reality Pet System Based on Passive Props . . . . . 645
Yaqiong Xue, Dongdong Weng, Haiyan Jiang, and Qing Gao

ReFall: Real-Time Fall Detection of Continuous Depth Maps

with RFD-Net. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 659
Yujia Zheng, Siyi Liu, Zairong Wang, and Yunbo Rao
xvi Contents

Parametric Display Surface Geometric Calibration Method Based

on Parameter Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674
Xiang Li, Shanshan Chen, Xiaoying Sun, Feng Chen,
and Dongdong Weng

An Optimization Method for Large Format Multi-projector Displays . . . . . . . 686

Xiang Li, Shanshan Chen, Dong Li, Xiaoying Sun, and Dongdong Weng

Study on Electromagnetic Visualization Experiment System Based on

Augmented Reality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 699
Xiaoxu Liu, Cong Wang, Jun Huang, Yue Liu, and Yongtian Wang

Research on Battle Damage Analysis Methods Based on Collaborative

Simulation for Armored Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 713
Jun-qing Huang, Wei Zhang, Wei Liu, and Tuan Wang

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725

 Human Fungal Infection Image Classification
Based on Convolutional Neural Network

Yuan Zhou(&), Yanxia Feng, and Haiying Zhang

Faculty of Automation and Information Engineering,

Xi’an University of Technology, Xi’an 710048, China
[email protected]

Abstract. An improved algorithm for deep learning of convolutional neural

network is proposed in this paper to automatically extract feature of the fungal
images. Firstly, the target image of the connected area is used to detect the
targets of the fungal image, and several small images of conidia in the original
image are obtained. Secondly, the small image is augmented by some opera-
tions, the augmented small images are proportionally divided into training sets
and validation sets, and the training accuracy and validation accuracy are
obtained. Finally, the test unknown images are input into the model, and the test
accuracy is obtained. Experimental results show that the measures of data
augmentation and fine-tuning not only effectively avoid the over-fitting of deep
learning algorithm in small samples, but also improve the accuracy. The training
accuracy of the algorithm can reach 95%, the validation accuracy can reach
96%, and the test accuracy can reach 69.23%, which has good robustness and
generalization.

Keywords: Fungal image classification
Data augmentation

Convolutional neural network (CNN)
Fine-tuning

1 Introduction

In recent years, with the aging of the population, organ transplantation, tumor radio-
therapy and chemotherapy, corticosteroid application and various catheter interven-
tions, the incidence of human fungal infections has increased year by year. From the
current epidemiology, human fungal infections can be divided into primary, secondary
and invasive. Their clinical manifestations are fever, pain and dyspnea, etc., because
they are not specific, combined with the complexity of the fungal spore microscopic
image itself and the similarity between different types of fungal morphology, so that
fungal infections Accurate and timely diagnosis and treatment have great difficulties.
At present, there are mainly the following methods for manually detecting and
diagnosing fungal types:
(1) Based on a combination of tissue culture and imaging. This method takes a long
time.
(2) Direct microscopy. The clinical specimens (sputum, bronchoalveolar lavage fluid)
were cultured and stained, then observed under a microscope [1].

© Springer Nature Singapore Pte Ltd. 2019

Y. Wang et al. (Eds.): IGTA 2019, CCIS 1043, pp. 1–12, 2019.
https://doi.org/10.1007/978-981-13-9917-6_1
2 Y. Zhou et al.

(3) Antigen examination based on GM experiment and G experiment. Diagnosis of

fungal infections by detecting antigenic components released by fungi into the
blood [2]. This method is of great significance for early diagnosis of fungal
infections, monitoring of high-risk populations and evaluation of curative effect.
(4) CT-based imaging examination. CT diagnosis is a more effective method com-
monly used in clinical practice. However, because CT images are mainly based on
one or two morphological performance, and other performance is combined, the
diagnosis is complicated and the infection category cannot be diagnosed [3].
In summary, the traditional method of artificial detection and diagnosis of fungi is
not only time-consuming and laborious, but the diagnosis results are easily affected by
many human factors. If computer-aided diagnosis (CAD) and artificial intelligence
methods are used to automatically classify microscopic images of fungal spores, not
only can the diagnostic efficiency be improved, but objective and accurate diagnosis
results can be provided [4].
The content of this thesis is that the deep learning method based on convolutional
neural network is adopted, and the data augmentation method and transfer learning are
adopted to prevent the over-fitting problem often encountered when the deep learning
algorithm is trained in small sample data sets. Thereby, the accuracy of microscopic
images of fungal spores is improved to meet the high standard clinical application
requirements.

2 Fungal Image Preprocessing

2.1 Fungal Conidia Target Detection

In Fig. 1, in a primitive fungal image, due to the existence of several conidums, the
whole image is input for training during the detection and classification process, and
the position distribution and density of each conidium in the image are obtained.
Background information such as size, bubble, and the like are extracted as features of
the image, and an over-fitting phenomenon occurs. Therefore, it is necessary to pretreat
each conidium in the whole image [5].

(A) Aspergillus terreus (B) Spore-forming structure of filamentous fungi

(a. conidia head b. spore stalk c. crest sac d. bottle stem e. sylph f. conidia)

Fig. 1. Original image of filamentous fungi and its sporulation structure.

 Human Fungal Infection Image Classification 3

In combination with the manual marking and the target detection method, the
training data is obtained by manual marking, and the test data is obtained by the target
detection algorithm, thereby automatically detecting the small target area in the test
image, that is, a single conidium region. The specific algorithm is:
Step 1: Convert a three-dimensional color image into a two-dimensional gray
image;
Step 2: Convert the gray image into a binary image according to the empirical
threshold;
Step 3: Perform expansion and corrosion on the binary image;
Step 4: Sort the size of the target connected domain;
Step 5: Calculate the coordinates and size information of the target center position
to determine the specific location of the conidia;
Step 6: The single conidium position information of the original image is detected
as several sub-images.
In the second step of converting the gray image into a binary image, the trans-
formed boundary threshold can be found by the maximum inter-class variance method.
The third step is mainly to reduce the amount of calculation and eliminate the influence
of small gaps and small holes on the calculation of the target connected domain. In
addition to considering this effect, the whole of the conidium is also considered in the
process of target detection with a tangent rectangle [6].

2.2 Data Augmentation

Deep learning usually requires a large amount of data to train, at least a few hundred
thousand images. A small amount of data often causes over-fitting and non-
convergence. But in reality, a large number of labeled medical fungal images are
expensive and difficult to obtain, so you need to “create data”. In this paper, the data
augmented method is adopted based on the limited labeling fungal images provided by
the cooperative hospital. The specific operations include: rotate, affine transformation,
scaling, mirroring in horizontal and vertical directions, contrast transformation and
their combined operation. The data set has been expanded by 33 times.
When using any angle of rotation, considering the complexity of the background of
the filamentous fungus image, when the data is augmented by any angle rotation, the
image size changes and a blank background is generated. Therefore, the difficulty in
enhancing the image data of the filamentous fungus is after the rotation. Blank back-
ground fills the problem.
As shown in Fig. 2, the original image size is m  n, the angle from the horizontal
plane is a, the distance from the center point to any corner point is q, and the position
coordinate of one pixel point is ðx0 ; y0 Þ, the image is rotated by h degrees, and ðx1 ; y1 Þ
is the pixel coordinates after rotate. The area of the circumscribed rectangle that is
tangent to it after rotation changes, so the canvas needs to be reselected to store the
rotated image matrix. In order to avoid information loss, the principle of canvas
expansion is to carry all the picture information in a minimum area. The smallest
canvas size selected is M  N.
4 Y. Zhou et al.

Fig. 2. Schematic diagram of pixel rotation coordinates

The horizontal and vertical coordinates after rotation are as follows:

x1 ¼ q cosða þ hÞ ¼ qðcos a cos h  sin a sin hÞ

ð1Þ
¼ x0 cos h  y0 sin h

y1 ¼ q sinða þ hÞ ¼ qðsin a cos h þ sin h cos aÞ

ð2Þ
¼ x0 sin h þ y0 cos h

M ¼ m cos h þ n sin h ð3Þ

N ¼ m sin h þ n cos h ð4Þ

x1 ¼ ðx0  m=2Þ cos h  ðy0  m=2Þ sin h þ M=2 ð5Þ

y1 ¼ ðx0  m=2Þ sin h þ ðy0  n=2Þ cos h þ N=2 ð6Þ

The matrix transformation formula is obtained by the combination of Eqs. (1), (2)
and (3) and (4):
2 3
cos h sin h 0
½x1 y1 1 ¼ ½x0 y0 1
4  sin h cos h 05 ð7Þ
0 0 1
2 3
cos h  sin h 0
½x0 y0 1 ¼ ½x1 y1 1
4 sin h cos h 05 ð8Þ
0 0 1

According to the matrix transformation formula in Eq. (7), the pixel position after
the image is rotated can be easily calculated. However, considering the coordinate
transformation, the rotated coordinate points are not necessarily integers, and there
must be some coordinate points in the new image. No corresponding pixels can be
 Human Fungal Infection Image Classification 5

filled, or the position arrangement destroys the original neighbor relationship. The
resulting rotated image will appear “empty”. It can also be considered as regular noise.
As shown in Fig. 3(a).

(a)"Void" phenomenon (b)reverse mapping (c)background filled

Fig. 3. “Void” phenomenon, reverse mapping and background fill after rotation

In order to solve this problem, a reverse mapping method, such as Eq. (8), it is
adopted to take the color forward. Starting from the rotated image, find the point of the
corresponding original image, and then pass the gray value in the original image, so that
each pixel of the rotated image can certainly correspond to a point in the original image,
Different strategies can make pixels more accurate. The effect is shown in Fig. 3(b).
Finally, for the background filling of the black background area in Fig. 3(c),
considering the difference in the background color of the original image is large, it is
not easy to directly assign the value, which is also the difficulty in enhancing the image
data of the filamentous fungus. After many experiments, the background mean value at
the width of 1/4 of the edge is selected to ensure that the filling result is close enough to
the background of the adjacent edge, and the visual difference of the transition region is
reduced.

3 Convolutional Neural Network

Convolutional Neural Network (CNN) is a feedforward neural network with deep

learning function. Its structure includes input layer, convolution layer, pooling layer,
fully connected layer and output layer. Its feature learning process including low-level
to high-level feature extraction and classification. The network can recognize two-
dimensional graphics of displacement, scaling and other forms of distortion invariance.
During the training process, it can be implicitly learned from the training data, and can
be learned in parallel [7], so in the field of image detection classification, there are a
wide range of applications.
6 Y. Zhou et al.

3.1 The Specific Structure and Design of Convolutional Neural Networks

(1) Convolutional layer: it’s responsible for the learning of features. The calculation
formula is as shown in (9).
" #
ðlÞ
X ðl1Þ ðlÞ ðlÞ
qj ¼f qi  kij þ bj ð9Þ
i2M l1

(2) Pooling layer: it’s responsible for aggregation statistics of adjacent area features.
The map of the convolutional layer is downsampled by taking the maximum or
average value of the region to reduce the input size of the next layer, thereby
reducing the number of parameters and the amount of calculation of the network.
(3) Fully connected layer: Pull the input into a column vector and multiply it by a
weight to get the classification result.
Assume that a picture is input, indicating that the label corresponding to the input
image is, where is the number of classification categories. For a given test image, the
hypothesis function is to estimate the probability value it belongs to each category.
2 3 2 T 3
pðyi ¼ 0Þjqi ; b eb 0 q i
6 pðyi ¼ 1Þjqi ; b 7 1 6 ebT1 qi 7
hh ð qÞ ¼ 6
4 pðyi
7¼P 6 7 ð10Þ
¼ 2Þjqi ; b 5 k1 hT qi 4 bT2 qi 5
e i e
i¼0 T
pðyi ¼ 3Þjqi ; b eb 3 q i
P
In the formula, k1i¼0 e
bTi qi
represents the normalization of the probability distri-
bution, the sum of the corresponding probabilities is 1. b represents the parameters of
the classifier, and Y represents the correct classification result. The loss function is

X
k1
e bi q i
T

Lossðq; Y; bÞ ¼  Yi  log Pk1 T ð11Þ

bi qi
i¼0 i¼0 e

Finally, the loss function is minimized according to the stochastic gradient descent
method to achieve the optimal classification purpose.
Using two convolutional neural network models for comparison, in addition to the
input layer and the output layer, the improved GoogLeNet network has 11 hidden
layers which are 4 layers of convolution, 4 layers of pooling and 3 layers of full
connections respectively [8–11]. Because of the difference in the structure of the two
models, the classification performance of the two models is different. In this paper, two
kinds of networks are used to detect and classify filamentous fungal images.

3.2 Fine-Tuning Training

After the data is augmented, the requirement to fully train a network is not achieved, so
in order to avoid over-fitting, transfer learning is adopted. There are four main transfer
methods of transfer learning: sample transfer, feature transfer, model transfer, and
 Human Fungal Infection Image Classification 7

relationship transfer. The sample transfer is mainly applied to the problem of small
target data samples effectively when the source data is very similar to the target data
samples. For example, the traditional Adaboost algorithm is promoted, Dai et al.
propose a boosting algorithm with transfer capability [12]. The idea of feature transfer
is to find similar feature representations between source data and target data samples,
such as CoCC algorithm, TPLSA algorithm and self-learning algorithm. Model transfer
is to map the target data into the source data feature space, and use a unified model for
learning and classification [13, 14]. For example, transfer learning from language
recognition to image recognition [15].
Considering that the image similarity between the filamentous fungal image and the
source data is not high, the model has little correlation as well, but it has certain
similarity with the source data features. For example, the characteristics of dandelion in
the ImageNet dataset are similar. Therefore, the feature transfer learning method is used
to retrain the filamentous of small samples to achieve the purpose of over-fitting [16,
17]. Feature transfer first pre-trains the network on the big data set, obtaining the
parameters of each layer of the network, and then copying its first n-layer parameters to
the first n-layer of the target network, freezing the first n-layer parameters, and the other
layers are randomly initialized. The specific operation process is:
Step 1: Use to train on the big dataset ImageNet the network and get the network
parameters.
Step 2: Freeze the first n layer parameters in the network, and other layers are
randomly initialized.
Step 3: Retrain the small dataset filamentous fungus image with the parameters in
step 2 as the initial parameters. Get training accuracy and validation accuracy.
Step 4: Change the number of layers frozen in the network, cycle step 2 and step 3,
find the optimal value of the number of transfer layers.
The purpose of transfer learning is to find the model parameters shared in the
ImageNet dataset and a small number of filamentous fungal images, and to optimize the
parameters of the filamentous fungal image classification model by using the source
model parameters, so as to realize the characteristic transfer of data in different fields,
which is beneficial to the filamentous fungal image. Classification

4 Results and Analysis

4.1 Data Augmentation and Target Detection

The collected labeled fungal image data set was used as training data and validation
data. The data set contained a total of 1128 labeled fungal images, including 198
Aspergillus nidus images, 440 images of Alternaria alternata, and Aspergillus terreus
320, 170 other fungal images. Each fungus image uses four different magnifications
(40X, 100X, 200X, 400X) and a fixed RGB three-channel image of 3024  4032
pixels. The augmented data set has been expanded 33 times with nearly 40,000 fungal
images.
8 Y. Zhou et al.

According to the target detection algorithm steps, three sets of target detection
experiments were carried out on the Aspergillus terreus image. The experimental
results were taken as the average of the multiple experimental results. The experimental
precision has good stability. One set of experimental results is shown in Fig. 4.

(a) Original image (b) Target test result

Fig. 4. Target detection effect of Aspergillus terreus image

It can be seen from Fig. 5 that most fungal conidia are detectable, but some targets
are still not detected, and there are leakage and misdetection phenomenon. The target
detection algorithm was tested on three Aspergillus species samples, and the test results
are shown in Table 1.

Table 1. Statistics on spore image detection results of Aspergillus terreus

Data set Aspergillus
terreus
a b c
Total spores 36 26 8
Number of detections 38 26 8
Number of misdetections 4 6 3
Number of missed detections 2 6 3
Accuracy (%) 94.4 76.9 62.5

As can be seen from Table 1, the accuracy of target detection is related to factors
such as the total number of detections and background complexity. The more the total
number of spores, the more concentrated the target distribution, and the less the
background interference information, the easier the computer can detect the target.

4.2 Fine-Tune Training Results and Analysis

On the basis of data augmentation, the transfer learning strategy is used to train the
network. In order to verify the effect of parameter transfer, the convolutional layer
 Human Fungal Infection Image Classification 9

learning parameters are first frozen, the full connection layer of the source model is
replaced with the 4-class Softmax classifier, and then the pre-training is performed. The
model is put into the concentration training of the filamentous fungus samples, that is,
the feature extraction ability and the induction ability of the pre-training model are
retained. In the experiment, the number of training iterations is 100, the number of
batches is 32, and the learning rate is 0.001. The results are shown in Table 2.

Table 2. Softmax classifier classification performance

Network Training accuracy/% Validation accuracy/% Time/min Parameter amount
GoogLeNet 93.1 82 53 10394614
Alexnet 74.42 71.64 32 58612952

Table 3. GoogLeNet network performance under different training parameters

Network Training Validation Number of Trained parameter
accuracy/% accuracy/% freeze layers ratio/%
GoogLeNet 91 88.76 5 96.1
92.59 90.66 4 93.98
98.22 96.55 3 75.4
94.94 94.84 2 65.66

It can be seen from Table 2 that after replacing the fully connected layer with the
Softmax classifier, the training accuracy and validation accuracy of the GoogLeNet
network is 10.36%–18.68%, higher than that of the Alexnet network, which is sig-
nificant. The training accuracy and validation accuracy of GoogLeNet network can
reach 93.1% and 82%, but the validation accuracy of Alexnet network can only reach
71.64%, which is far from meeting the clinical requirements. In order to further
improve the network performance, the model parameters are fine-tuned, and different
model structures use different frozen layers. Other parameters are unchanged, and the
network performance changes under different trainable parameters are observed, and
the number of frozen layers and training can be found. The optimal value of the
parameter scale. The results are shown in Tables 3 and 4.

Table 4. Alexnet network performance under different training parameters

Network Training Validation Number of freeze Trained parameter
accuracy/% accuracy/% layers ratio/%
Alexnet 80.23 79.69 4 96.9
84.56 83.2 3 96.17
84.14 82.42 2 34.25
78.92 75.39 1 6.83
10 Y. Zhou et al.

It can be seen from Tables 3 and 4 that the optimal validation accuracy of the two
networks can reach 96.55% and 83.2%, and the network performance is greatly
improved, indicating that the fine-tuning strategy has a significant improvement effect
on the network classification. When the two networks are frozen at 3 layers, that is, the
GoogLeNet network trainable parameters account for about 75%, and the Alexnet
network trainable parameters account for about 96.3%, the network performance is the
best. The GoogLeNet network training accuracy and validation accuracy can reach
98.22% and 96.55% respectively, which can provide reference for the clinical detection
of filamentous fungi. The performance of the two network classifications gradually
become better as the number of frozen layers increases, but when it increases to the
optimal value, the network performance gradually deteriorates. As shown in Fig. 5.

Fig. 5. Alexnet network different frozen layer accuracy changes

The performance of the Alexnet network increases first with the number of frozen
layers. When the optimal value reaches reached, the performance begins to decrease.
The optimal number of frozen layers is 3 layers, and the validation accuracy is up to
83.2%. After the transfer learning, the performance has improved. The main reason is
the network structure is simple, and the parameters that can be migrated with a small
correction range. Therefore, the accuracy is low, but the Alexnet network runs faster
than the GoogLeNet network.
Figures 6, 7, 8, and 9 show the change of the accuracy and the loss value of the
training process of the GoogLeNet and the Alexnet network transfer learning model. It
can be seen from Figs. 6 and 7 that the GoogLeNet starts to converge 10 times in
iteration, and the difference between the training precision and the verification accuracy
is small, and there is no over-fitting phenomenon. The loss values of the two are close
to each other in the iterative process, and finally can be reduced. By 7–10, the highest
accuracy is around 95%. In Figs. 8 and 9, the Alexnet network starts to converge 20
times in iteration, but the convergence effect is not good. There is a difference between
the training accuracy and the verification accuracy. There are multiple oscillations, and
the final loss value is 10–15. The optimal accuracy is about 85%. This result indicates
that the multi-size convolution kernel design in the GoogLeNet makes its parameter
utilization higher and generalization better. The main reason for the poor performance
of the Alexnet network classification is that the training dataset is too small.
 Human Fungal Infection Image Classification 11

Fig. 6. GoogLeNet classification performance Fig. 7. GoogLeNet training process loss value

Fig. 8. Alexnet classification performance Fig. 9. Alexnet training process loss value

References
1. Li, Y., Su, H., Qi, C.R., et al.: Joint embeddings of shapes and images via CNN image
purification. ACM Trans. Graph. 34(6), 1–12 (2015)
2. Gu, J., Wang, G., Cai, J., et al.: An empirical study of language CNN for image captioning.
Comput. Vis. Pattern Recogn. (2017)
3. Gyvez, J.P.D., Wang, L.W.L., Sanchezsinencio, E.: Large-image CNN hardware processing
using a time multiplexingscheme. In: Fourth IEEE International Workshop on Cellular
Neural Networks & Their Applications. IEEE (1996)
4. Wang, W., Yang, L.J., Xie, Y.T., et al.: Edge detection of infrared image with CNN_DGA
algorithm. Int. J. Light Electron Opt. 125(1), 464–467 (2014)
5. Li, G., Yu, Y.: Visual saliency detection based on multiscale deep CNN features. IEEE
Trans. Image Process. 25(11), 5012–5024 (2016)
12 Y. Zhou et al.

6. Zhang, B., Wang, L., Wang, Z., et al.: Real-time action recognition with deeply-transferred
motion vector CNNs. IEEE Trans. Image Process. 27, 2326–2339 (2018)
7. Cao, J., Pang, Y., Li, X.: Learning multilayer channel features for pedestrian detection. IEEE
Trans. Image Process. 26(7), 3210–3220 (2017)
8. Liu, P., Guo, J.M., Wu, C.Y., et al.: Fusion of deep learning and compressed domain features
for content based image retrieval. IEEE Trans. Image Process. 26, 5706–5717 (2017)
9. Kn, L., Ravindhranathreddy, B., Suryakalavathi, M.: Brain storm optimization algorithm for
solving optimal reactive power dispatch problem. Chronexus.Org (2014)
10. Sun, Y.: A hybrid approach by integrating brain storm optimization algorithm with grey
neural network for stock index forecasting. Abstr. Appl. Anal. 2014, 1–10 (2014)
11. Wang, J., Hou, R., Wang, C., et al.: Improved v, -Support vector regression model based on
variable selection and brain storm optimization for stock price forecasting. Appl. Soft
Comput. 49, 164–178 (2016)
12. Tajbakhsh, N., Shin, J.Y., Gurudu, S.R., et al.: Convolutional neural networks for medical
image analysis: fine tuning or full training? IEEE Trans. Med. Imaging 35(5), 1299–1312
(2017)
13. Affonso, C., Rossi, A.L.D., Vieira, F.H.A.: Deep learning for biological image classification.
Expert Syst. Appl. 85(11), 114–122 (2017)
14. Smith, R.: The 7 Levels of Change, 2nd edn. Tapeslry Press, Littleton (2002)
15. Wang, S., Liu, L., Duan, L., et al.: Accurate segmentation of Ulva prolifera regions with
superpixel and CNNs. In: International Conference on Security. IEEE (2018)
16. Richter, C.D., Samala, R.K., Chan, H.P., et al.: Generalization error analysis: deep
convolutional neural network in mammography. In: Computer-Aided Diagnosis (2018)
17. Jang, J., Kwon, J.Y., Kim, B., et al.: CNN-based estimation of abdominal circumference
from ultrasound images (2017)
 Improved RPN for Single Targets Detection
Based on the Anchor Mask Net

Mingjie Li(&), Youqian Feng, Zhonghai Yin, Cheng Zhou,

and Fanghao Dong

Foundation Department, Air Force Engineering University, Xi’an, China

[email protected]

Abstract. Common target detection is usually based on single frame images,

which is vulnerable to affected by the similar targets in the image and not
applicable to video images. In this paper, anchor mask is proposed to add the
prior knowledge for target detection and an anchor mask net is designed to
improve the RPN performance for single target detection. Tested in the
VOT2016, the model perform better.

Keywords: Anchor mask
RPN
Single targets detection
Timing

Time series

1 Introduction

Since the introduction of deep learning technology into the field of computer vision, all
tasks for single-frame images are done well by a variety of networks. In the field of
image recognition, Alexnet, VGG, Resnet [1] and more constantly refreshed the correct
rate record of the Imagenet game. In the field of Target segmentation, FCN [2] and
perfect application of the technology of CRF [3, 4] on it make the segmentation effect
is more and more significant. AND in the field of Target Detection, the series of RCNN
[5–7], the series of YOLO [8–10], R-FCN [11] and SSD [12] perform better and better
by constantly improving the structure. But the target detection on the singe image
frame can be affected simply if there is no prior knowledge added in the network. And
with the development of target detection networks, the basic structure of RPN in also
used in the all mainstream algorithm (for instance Faster-RCNN, Yolo-v3, and SSD).
According to the problem and character of RPN, the concept of anchor mask is pro-
posed in this paper, which can be used to connect to the prior knowledge to filter a lot
of wrong anchors to improve the detection accuracy. An anchor mask net considering
the timing characteristic is designed to join to the RPN as an auxiliary part. Passing the
test in the VOT2016 data set, new RPN preforms better for single targets detection.

2 Relation Work

2.1 RPN
RPN is proposed firstly in Faster-RCNN, which generate several anchor boxes with
different scales and ratios according to the anchor points on feature maps. And anchor
© Springer Nature Singapore Pte Ltd. 2019
Y. Wang et al. (Eds.): IGTA 2019, CCIS 1043, pp. 13–19, 2019.
https://doi.org/10.1007/978-981-13-9917-6_2
14 M. Li et al.

boxes can be adjusted to location the target and provided confidence scores by CNN.
RPN is a effective structure to get proposal boxes which replace the complex procession
of research boxes generation in RCNN. The structure of RPN is as showed in Fig. 1.

Fig. 1.

In this paper, a RPN is trained to be a target detector for video clip frames.
Because RPN is the basic structure for the main steam algorithm in the targets detection
field. So our anchor mask net will be useful in other networks if it is fine to RPN.

2.2 FCN
FCN is a classical structure for target segmentation. It gets the image features by the
convolutional layers and pool layer firstly. After that, resize the feature maps to the
same size of the original image by the fully convolutional layers and de-convolutional
layers to predict the classes of each pixel.
In this paper, a similar FCN is trained to predict the valid anchor points. The
structure of FCN is shown in Fig. 2.

Fig. 2.

2.3 3dCNN
Three-dimensional Convolutional neural Networks (3dCNN) is firstly use for video
analysis proposed in the C3D [13]. This structure integrates temporal and spatio
information of video. Similar to ordinary convolution operations, 3dCNN make clip
frames at different times equivalent to channels in 2dCNN. Convolutional operation is
performed in each time dimension, and then an addition operation is performed on all
time channels. In this paper, a 3dCNN is used to be the front end to get time infor-
mation of the IOU heat maps of the first three frames. The principle of 3dCNN is
shown in Fig. 3.
Discovering Diverse Content Through
Random Scribd Documents
march to headquarters. I asked one of the privates, did he know the
Corporal. "He joined about three weeks ago, sir."
"Hope he's well. Where is he now?"
"He's there, sir," said the man, pointing to the parapet of the
redoubt. I looked, but saw no one.
"The earth," said Gingham, "seems to have been recently stirred
there. That mound, I think, is not many days old." Then, addressing
the soldier, "your regiment suffered a heavy loss. Is that where you
buried after the action?"
"That's the place, sir." The man then walked away, as if little
disposed for conversation.
We did not pause to calculate how many bodies would fill a space
commensurate with the length, breadth, and altitude of the soil
displaced. There lay the slain of a gallant regiment, in the redoubt
they had so nobly won. There lay Corporal Fraser, who, in all the
difficulties of our march, had shown himself trustworthy, fearless,
intelligent, and energetic. He had longed to join ere the day of
combat, and had found a soldier's grave.
We discovered at length the sergeant who had informed me of my
cousin's wound. He now pointed to a large house, near the thicket at
the bottom of the hill. While searching for Cousin Tom on the day of
the fight, I was close to that same house, but without seeing it.
From our present elevated position it was distinctly visible, though
not from the low ground, amongst trees and underwood.
Our approach to the house led us through the thicket. While
making our way among the trees, we both, Gingham and I, came to
a halt at the same instant. The sight which arrested our steps was
new to Gingham, not to me. I saw, on that spot, an object that I had
seen two days before. The sergeant whom I had then found
wounded was still sitting there, on the same bank, in the same
attitude! There he had sat the whole time, overlooked by the
bearers, and unable to move. Viewed at the distance of a few paces,
his aspect scarcely appeared changed. It was the identical figure—I
remembered him at once. But on a nearer inspection, the alteration
was but too manifest. His eye was glazed, and half shut. His face
was that of a corpse. He sat up, like a dead man galvanised. "What,
still here, sergeant? Has nobody come to remove you yet."
He attempted to speak—paused—at length found utterance.
"Sorry I didn't accept your offer, sir." His voice was low and husky,
but distinct.
"Come," said Gingham, "you mustn't refuse this time. We'll soon
carry you into the house just by."
"Thank you, sir; thank you, gentlemen. Would you have the
kindness though—I should be sorry to lose my gaiter."
The gaiter secured, we prepared to lift the sufferer from his seat,
and he on his part made a feeble effort to rise. The attempt brought
on a gush of pain. For a moment, his features were distorted with
intolerable anguish; the next, he fainted in our arms.
"Now then," said Gingham, thrusting back into his sidepocket a
small flask which he had just drawn out. "Now then; away with him
at once, before he recovers. Come, Mr Y——; you take his shoulders,
I'll take his legs. It may save him further pain."
We bore the sufferer, still senseless, to the house. Gingham, not
having a hand to spare, banged at the door with his foot. It was
opened by Mr Staff-surgeon Pledget, who bowed on recognising us,
but looked rather perplexed at the unexpected addition to his duties.
Pledget gave instant directions for the accommodation of the
wounded man, and informed me, in reply to my inquiries for Cousin
Tom, that he had an officer under his care, answering to my
description. Pledget appeared bewildered, and stood with us in the
passage a few moments, without speaking. At length he opened the
door of a small chamber close by, and begged us to enter. He placed
chairs for us, and seated himself on the bed. "I'm rather exhausted,"
said he.
 "I fear after such a fight," said Gingham, "your duties must be
heavy indeed."
"Oh yes," said Pledget, looking distressed and rather wild. "I have
had much work, and little assistance; a long spell, too."
"Why, you began, I suppose," replied Gingham, "early on the day
of the fight."
"Yes," said Pledget; "and I've been at it ever since. Let me see:
two days and two nights, isn't it? Yes, and now going on for the
third. Here have I been operating, bandaging, taking up arteries,
taking off arms and legs, night and day, without time to lie down,
almost without a moment to eat. In fact," said he, looking about the
room like a man lost, "this is the first time I've sat down these eight-
and-forty hours."
Pledget's look bore full testimony to his toils. Three weeks' illness
could hardly have wrought a greater change. Nor was his
appearance mended by his garb. He wore a sort of operating gown
similar to that employed in dissecting; a long pinafore with sleeves,
protecting the whole person from the chin to the feet, tied round the
middle, and closing with a fold behind. The front was spotted in
every part with jets of blood from wounded arteries. Some of the
stains had dried on, and blackened where they dried; others, more
recent, were still moist and crimson. Blood was on his unshaven and
haggard face; and on his hands, too, wore marks of blood.
Gingham eyed him with a look of deep concern. "I really fear,"
said Gingham, "you've been quite overdone."
"I did hope, before this," replied Pledget, "to be relieved by other
gentlemen of my own department. I have but one medical assistant,
and he, at this moment, can afford me no help, for I have been
forced to leave him sitting with his finger on a wounded artery; and
if he takes it off but for a few seconds, the major's a dead man."
Pledget now looked like a man that can't remember what's next.
"Oh," said he, in all absent tone, "so peace is really concluded.
Come, Mr Y——, suppose we go and look for your cousin. His case,
I'm happy to say, is not serious. The ball will be extracted this
evening, and then, I hope, he will do well."
Pledget spoke, but did not stir. "By the bye," he added, "you know
Captain Gabion? I think you do. Oh yes, I recollect; we were all
three fellow-passengers from Lisbon to Falmouth. No, no, what am I
saying? From Falmouth to Lisbon. His case is past hope. He can
hardly live through the night."
Gingham and I rose at once from our seats. For the moment, the
imminent danger of a man we so highly esteemed, expelled from my
thoughts even Cousin Tom. Pledget also rose, as if to lead the way,
but again lapsed into forgetfulness. His mind was evidently worn
out, as well as his body. "Well," said he, "I'm glad we've got
Toulouse.—Gentlemen, I beg your pardon. This way, if you please;
up stairs."
He led the way. Every open door, as we passed through the
spacious mansion, discovered a room crowded with wounded and
dying men, in beds, or on the ground. Or, if we saw not into the
apartment, sounds were heard, which told of anguish and laceration
within. We were conducted by Pledget into a large room on the first
floor, filled, like the others, with every form of suffering. Some,
slightly wounded, sat round the fire, on which cookery was
proceeding in kettles of every size and shape. One officer, bandaged
round the head, had become delirious. He alternately laughed and
whimpered, muttered and sang. Another sat near him, moaning,
with his arm in a sling. A spent cannon ball had smashed the bones
from the elbow to the wrist, without inflicting an external wound.
Every bed had an occupant; and many lay upon the floor, with only a
blanket under them. My eye glanced round the apartment, and
lighted on the pinched features and pallid visage of Captain Gabion.
He lay on his back in bed. Death was legible in his aspect. His
eyes were all but shut; but, from time to time, a convulsive twitching
of the muscles suddenly expanded them to their full width. To all
appearance, he was perfectly insensible. His breathing was irregular
and laborious; but the expression of his countenance, except when
disfigured by the spasms which occasionally shot through his frame,
and jerked him from head to foot, was, as in health, calm and
dignified. Strange indeed were the vicissitudes, strange was the
contrast, between the rigid tranquillity of one moment, and the
awful distortion of the next. Now, it was the quivering play of
features pulled by muscular contraction; now, the monumental
repose of marble.
"I fear," whispered Gingham to Pledget, "you view the case
unfavourably." Pledget hopelessly raised his eyes.
"The Captain has been insensible," said Pledget, "ever since he
was brought in; and probably will continue so till he expires."
We turned from this sad spectacle, without exchanging a syllable.
A handkerchief was whisked in my face. I looked round; there was
the man I wanted. In the next bed, tucked in, with smiling face, little
changed since we parted, a splendid specimen of the ugly-handsome
—those fellows that make the biggest holes in ladies' hearts—lay
Cousin Tom. Gingham, my object attained, forthwith took a
temporary leave—had urgent business in Toulouse—an appointment
—would return as soon as possible.
"Fine fellow, that" said Cousin Tom, craning round, and nodding at
Captain Gabion.
"Well, Tom," said I; "what's the matter with YOU? What brought
YOU here?"

"Oh, not much; nothing," replied Tom, curling out his lips
contemptuously, like a disappointed man; "only a musket-shot. It
won't get me a step, I'm afraid; no, nor a pension neither."
"Well, but how was it? When was it? We lost you in a moment."
"I'll tell you just how it happened," said Tom. "You saw the old
colonel knocked over. Ah! Don't touch the bed; that's a good fellow.
Well. Directly after, you know, we charged. I was running on; felt a
smart crack in the small of my leg. Thought it was a stone; took no
notice. A few paces further, though, found I couldn't walk. The
sergeant looked at my leg; said 'You're wounded, sir.' Wounded I
was, sure enough; and disabled, too. Got carried to the rear; placed
myself in the doctor's—"
Here Tom suddenly knit his brows. His colour changed in an
instant from florid to livid; his whole face was distorted with pain.
Clapping his handkerchief to his mouth, he chewed away at it with
all his might, while big drops of sweat started out on his forehead,
and he drew in breath till the bedclothes heaved. Next moment he
was himself again.
Once more Tom nodded at the next bed. "Known him long? The
doctor knows him."
"We came over from England, all three of us in a ship."
"Doctor was out, though, in one thing," said Cousin Tom. "Told
you he was insensible ever since he came in. No such thing; this
morning he revived; for about an hour seemed quite himself. Told
me how he got hit."
"Then tell me. I must communicate with his friends in England."
"Well," replied Tom, "the Captain wasn't on duty here at
headquarters; was doing some field-works on the left bank of the
river, to be ready for Soult in case of his bolting again for the south.
He heard, though, that the fight was coming off; so rode in on the
morning. Found out there was to be a flank movement to the left;
thought he might as well explore the line of march; went forward
alone. Passed through the thicket on foot; made his way from one
end to the other. When he reached the further extremity, just where
our men got such a pounding afterwards from the guns on the
heights, he looked out for the enemy's skirmishers; saw no one;
thought he might as well go a little further. Just then our batteries at
the right opened on the French position; some of our shots flew too
high, and came clean over the hill into the lane, just exactly where
he was standing."—Indeed! I thought of Captain Gabion's dream.
—"Well; he saw one coming; didn't trouble himself; it seemed spent.
Just when he thought it was going to stop, it fetched a pitch; took
him in the side. He was found when our troops advanced, and
brought in here." At this moment the pain returned. Tom again
made wry faces, took another chew at his handkerchief, and soon
recovered as before.
"Well, Tom; I'm a leisure man. What can I do for you? Is there
anything you want?—anything I can get you?"
Cousin Tom looked very much as if there was something he did
want, yet was backward to speak. "Why," said he, "I suppose by this
time you can get into Toulouse. I wish you would make inquiries; try
and find me some—But never mind; it's of no use. The ball will be
extracted this evening, and to-morrow I shall go in myself."
"Nonsense, nonsense; I'll go this instant."
"Don't be too sure of that, though," said Tom. "Yesterday morning
I tried it. Told the servant to have my mule ready; got my things on
while the doctor was sawing away on the other floor; slipped down
stairs; gave him the go-by. Mounted—rode to the top of the hill—
was riding down into the city—almost rode into a French piquet."
"No fear of that now, Tom; the city is ours. I saw the French
troops marching out. Come, tell us, old fellow. What is it you fancy?
Anything the doctor sanctions, you know. A quarter of mutton?—a
dozen of pigeons?—some prime French sausages?—a bushel or so of
oysters? What do you say to a brace of biddies?"
"Oh, no!—oh, no!" said Tom, as if the very mention of biddies
made him sick. "We were always in advance; got fowls and turkeys
till we hated the sight of them."
"Any dish from a French cuisine, then?"
"Oh, no—oh, no! Nothing French, nothing Frenchified. What I
want, if it's to be got at all, is not to be got good, except in England
—or the West Indies."
"Well, but, you know, Bordeaux is open; West India produce has
come into the country by ship-loads. What is it? Come, just tell us,
old chap, and I'll go and get it for you at once, if it's to be had in
Toulouse."
Tom was not so well as he looked; and there was evidently
something for which, like other sick persons, he was inwardly pining.
Now that I had held out a prospect of its attainment, his cheek
flushed, and his eye gleamed with feverish eagerness.
"Well, then," said Tom, "I wish you would try and get me—but it's
no use; it's a shame to bother you.—I say, though, can you spare
the time? Have you really nothing to do? Upon your honour?—I've
been longing for them, day and night, ever since I got here. Oh, if
you could only get me—some tamarinds!"
His eye, while he spoke, fixed full on mine. He watched my
countenance with the anxiety of a dying man when he makes his
last request. "I'll be off and try this instant," said I, though really
fearing there was little chance of success.
"Oh, thank you—thank you!" cried Cousin Tom. I was going. "Here
—here! Come back! I want to speak to you!" I returned. "Old
fellow," said Tom, with a coaxing, eager grin, "make haste now, will
you? Bring 'em directly—that's a good chap."
"Well, but, you know—if tamarinds can't be had for love or money,
is there nothing else?"
"No, stupid—no! Tamarinds, I say; get me some tamarinds. What
did I tell you? Didn't I tell you tamarinds? Now then; what are you
waiting for? Cut away, and be hanged to you! Be off!—be off!"
I entered the ancient and very interesting city of Toulouse, and
rushed through streets choked with cars of wounded men, in search
of tamarinds. The search was tedious, and far from satisfactory. I
inquired at all the likeliest shops; found only two where they
professed to sell tamarinds. The samples were similar: a made-up,
sticky mess; a black, nauseous electuary, with a beastly
pharmaceutical odour, and barely the flavour of tamarinds.
 It was no pleasant thought returning to poor Tom with a big
gallipot of this filthy compound stowed in each of my coat pockets.
Yet, though bad thus to baulk him, it was worse to keep him in
suspense; so I started on my return with all speed, and, in my
speed, came full butt against a passenger, who hugged me like a
wrestler, to prevent a mutual capsize.
"Well, Mr Y——! Glad to see you so active. Something of
importance, no doubt: official duty, I suppose."
It was Gingham! I told him my troubles, my pursuit in behalf of
Cousin Tom, and my disappointment. Had searched all Toulouse, and
could find no good tamarinds.
"Shall be happy to supply you," said Gingham, "in any quantity
your cousin can require. Got a whole kegful—capital. Always take
some with me when I visit the Continent. Got them on Fish-street
Hill." We walked off forthwith to Gingham's quarters.
I was speedily on my return to Cousin Tom, with Gingham for my
companion, and a good jar of prime, sweet, wholesome,
unsophisticated tamarinds. On approaching Tom's bed, I held up the
jar in triumph. Tom raised himself without saying a word, tucked his
handkerchief under his chin, and sat up, poor fellow, like a child,
with eyes half-closed and mouth half-open, eager to be fed. In went
a spoonful. The next instant—bolt!—it was gone! What a swallow!
He sat as before, ready for another. A second allowance vanished
with equal speed. Down it goes! Why, it's like feeding a young rook!
—Tom now laid himself down again, exhausted. "Here," said he; and
made me a present of a handful of tamarind stones. "Now put a
good lot in that jug, and fill it up with water."
While the drink was mixing, an unusual sound called our attention
to the adjoining bed. Captain Gabion was fast sinking. His
respiration, laborious from the first, had now become painfully
audible; in fact, he did not breathe, he gasped. The convulsive
movements had ceased. His face retained its natural expression; but
there was that in his look which told us he was a dying man. I felt at
the moment an impression,—He is not insensible! His lips moved.
Surely he is trying to speak! He strove to fix his eyes on us, but
could not. I stooped down, observing his lips again in motion. Yes,
he was speaking. I caught only the words—"On the platform."
"The Calvinet platform?" I whispered in reply. "Is that the spot
where you wish—?"
Feebly, tremulously he pressed my hand, which had just before
taken his. I had caught his last request, then; a grave on the summit
of Mount Rave, the key of the French position, where the table-land,
crowned with redoubts, had been carried by our troops. His
breathing became gradually feebler and less perceptible. The
moment when it ceased entirely, no one present could determine.
This only was evident:—a minute before, he had given signs of life;
and now, he had passed into another world!
Cousin Tom's bullet was extracted the same afternoon, with
immediate relief to the patient. During the operation I was present,
by Tom's request; and friendship, let me tell you, has more pleasing
duties than that of attending on such emergencies. Tom, however,
made it as agreeable as he could. Throughout the process he
viciously stared me full in the face, grinning most horribly from time
to time, half in agony, half in fun. When the forceps was produced,
he caught a glimpse of that terrific implement, and twisted his ugly
mug into such a comical grimace, that mine, spite of the solemnity
of the occasion, was screwed into a smile. Tom thereupon clenched
his fist, with a look that said ferociously, "Laugh again, and I'll punch
your eye."
The bullet, doctor, had lodged between the bones of the leg, a
little above the ankle, and, I need not inform you, came out rather
flattened. Tom kept it as a bijou, in a red morocco case made
express by an artist in Toulouse. Tom called it his pill-box. Neither
bone was broken; but the strain of this disagreeable visitant wedged
in between them, and rending them apart, had occasioned from time
to time those awful twinges, which Tom assuaged by taking a chew
at his handkerchief. The enemy removed, he not only found himself
in a state of comparative ease, but was relieved from the
constitutional irritation which had begun to manifest itself by
hardness of pulse, dryness of the mouth, parched lips, a dull, hectic,
brickdust-coloured patch on each cheek, a feverish lustre of the eye,
and an enormous appetite for tamarinds.
The operation, though, I ought to have said, was not performed
by Pledget, but by another army surgeon, who had arrived in the
course of the day, not before he was wanted. Poor Pledget was quite
done up. His powers, both mental and physical, had evidently been
over-taxed. He looked haggard and wild. Yet still, though relieved,
anxious about his cases, he wandered from room to room, and
fidgeted from one patient to another; standing a while in silence,
with his hands behind him, first by an amputation, then by a
wounded artery, then by a contusion, then by a broken head; while
his care-worn countenance expressed pleasure or pain, according to
the symptoms. As Cousin Tom was now in a dreadful fuss to be off
for Toulouse, Gingham and I applied to the newly-arrived surgeon,
and consulted him as to the removal.
"I think, gentlemen," said he, "if no bad symptoms supervene in
the night, it may safely be effected to-morrow; that is, of course,
with proper care and precautions."
"You are not afraid, sir," said Gingham, "that to-morrow may be
too early a day, then?"
"Why, sir, to say the truth," replied the doctor, "if we had more
room here, better accommodations, and a less vitiated atmosphere,
I should say a later day would be better. But, under existing
circumstances, less evil, I think, is likely to arise from the patient's
removal, than from his remaining. In his case, what we now have
most to look to, is the general health. Keep that right, and the
wound, I hope, will do well. Therefore the sooner he is withdrawn
from the bad air, and the associations which surround him here, the
better for him." The doctor paused.—"Pray, sir," said he, looking
Gingham full in the face, as though intuitively knowing he spoke to a
real good fellow, "pray, sir, if you will permit me to ask the question,
is Mr Pledget a friend of yours?"
"There are few men, sir," replied Gingham, "for whom I have a
higher regard, than for Mr Pledget."
"Well, sir," said the doctor, "I feel rather uneasy about him. It's a
delicate thing to speak about. But you yourself must have noticed
how changed he is, by the labours of the last three days. In short, to
speak plainly, he requires to be looked after; and just at this time,
with so many wounded upon our hands, I hardly know whether we
could possibly give him the attention here which his case requires. If
it is neglected now, it may become serious. Would it be asking too
much, if I requested you to take charge of him into Toulouse?"
"Take him with us this instant, sir," said Gingham; "or when you
please. If you approve, I'll have him with me in my own quarters."
"I really, sir, feel obliged to you," said the doctor. And the doctor
looked as if he spoke from his heart. "Hope you understand, though,
what it is you are taking on your shoulders. For a few days—not
longer, I hope—he will require vigilant superintendence, and,
possibly, slight control. His case demands firmness, and indulgence
at the same time."
"Yes, sir, I understand," said Gingham. "Shall he go with us now?"
"I would rather have him under my eye," said the doctor, "till to-
morrow morning. Perhaps a night's rest may effect a favourable
change. In the interval, too, I shall have time to prepare his mind for
the removal." So it was settled.
The next morning we returned to the chateau, for the purpose of
bringing in Pledget and Cousin Tom. Tom's patience, though, had not
lasted out till our arrival. At sunrise, again giving the doctor the go-
by, he had got on his things, crept down stairs, mounted his mule,
and taken himself off. In fact, he had got into Toulouse, obtained a
billet, and, snugly located in a respectable French family, was
prattling the vernacular, which he had at his fingers' ends, before we
arrived at the chateau to fetch him.
 It only remained, therefore, to remove Pledget. He, poor man,
though all the better for a night's rest and a clean shirt, still looked
very unlike himself. He had rested, indeed, but he had not slept; and
his medical colleague hinted to Gingham, ere we departed, that the
case still required vigilance and care. The state of Pledget's mind, at
this time, was singular; he had all at once become excessively
ceremonious. When we reached the garden gate he drew up;
insisted that we should both precede him in going out. Had Gingham
and I been equally punctilious, we should not have reached Toulouse
by dinner-time.
Gingham had a matter upon his mind. Captain Gabion having
expressed a last wish respecting his funeral, Gingham had
undertaken the whole details, and some arrangements had been
necessary at the chateau, or our departed friend would speedily
have been consigned, on the spot, to a ready-made grave. Gingham
mentioned the subject as we rode along, and began stating what
steps he had taken. Pledget, who was ambling side by side with us
on his mule, suddenly fell behind. Coosey, previously admonished by
Gingham, kept still further in the rear. We waited till Pledget came
up.
"Why, Mr Pledget," said Gingham, "I thought we had lost you, sir."
"Excuse me, sir," said Pledget, with gravity; "you are making a
confidential communication. Part of it I unintentionally overheard.
For this, an apology is due to both of you. Gentlemen, I most
humbly beg your pardon."
We rode on. Presently, Pledget edged up alongside of me, as
though he had something important to communicate.
"Mr Y——," said he, "I consider it the first duty which one
gentleman owes another, to avoid giving him needless offence." Not
exactly perceiving to what this observation tended, I could only bow
my acquiescence.
"But if," continued Pledget, "an offence is actually given, then I
conceive the next duty is to make reparation by a humble apology."
Apology, it was evident, was now the uppermost idea in poor
Pledget's mind.
"Well, sir," said I, seeking to divert his thoughts, "I think, in such a
case, regard should be had to the feelings of both parties. And,
judging by my own, I should say that, next to making an apology,
there are few things one would more wish to avoid than receiving
one."
"And accordingly," said Gingham, "in the intercourse of gentlemen,
it rarely, very rarely occurs, that an actual apology is deemed
requisite. To signify an intention, to express a willingness to
apologise, is in most cases thought amply satisfactory. Manly feeling
forbids the rest; and honour itself exacts no more." Pledget rode on
awhile, absorbed in thought.
"Mr Y——," he said at last, "I appreciate your sentiments, as well
as Mr Gingham's; and I perceive their drift. Allow me to say it, your
conduct is most generous. I really feel that you have just cause to
complain of mine; and, if it would pain you to receive the apology,
which is your due, allow me at least to express my willingness, and,
believe me, it was my intention, to apologise."
"Mr Pledget, my dear sir, what possible need of apology between
you and me? What offence has been given or received? I know of
none—never dreamt of any."
"Very handsome of you to say so, Mr Y——," replied Pledget. "But
what could be more inconsiderate than my conduct yesterday
morning? You must have felt it; I know you did. You came to me
with an anxious inquiry respecting your wounded cousin; I spoke to
you of Captain Gabion. It was wrong, I own. Nay, not merely wrong,
it was unfeeling. I trust you will bear in mind my peculiar
circumstances at the time. I was overwhelmed, perplexed,
bewildered, I——"
Gingham now saw it was high time to interpose, and with much
adroitness gave a new turn to the conversation. But ere we were
housed in Toulouse, Pledget, addressing us alternately, and
continually discovering fresh grounds of self-accusation, had made
two or three more apologies.
For a few days, sedulously and most kindly tended by Gingham,
who managed him admirably, and evinced equal tact and delicacy,
Pledget continued in a state of alternate depression and excitement,
with occasional hallucinations. He made apologies to all who came
near him; and, ere he quitted Gingham's quarters, had begged
pardon, again and again, of every servant in the household. From
my first conversation with Gingham on the steps of the hotel at
Falmouth, I always valued his acquaintance. But when I had seen
him in this his new character as Pledget's nurse, wise, thoughtful,
vigilant, and indulgent, I really grew proud of such a friend.
Within a week Pledget was almost himself again; and long before
he quitted Toulouse, to embark for England at Bordeaux, he was
fully and permanently restored.
Cousin Tom's, though, was a business of more time. He begged or
borrowed a formidable sapling, with a knob as big as his fist, and
was soon able to hobble about Toulouse, very much to his own
satisfaction. But the bones of his leg had been injured, though not
broken; and it was long before the wound got well, if it ever did. I
was with him many months after in London, when the Medical Board
sat to award gratuities and pensions to the wounded and disabled
officers of the Peninsular Army. Lucky, then, did the wight esteem
himself who had lost a limb or an eye. Tom was waiting for his turn
to go before the Board; I saw him two days previously. His, I feared,
was only a case for a gratuity; but Tom was determined to go for a
pension, and made sure of getting it. I ventured to express my
doubts; Tom whipped off his half-boot, turned down his sock, and
exclaimed triumphantly, "Look at that!" The wound was clean, but
looked fresh; much, indeed, as it appeared two days after the fight
when the bullet was extracted, and still big enough to re-admit it. "If
the Board don't give me a pension," cried Tom, "for such a punch as
that, why, all I can say is, they deserve to be punched themselves."
Saw him again after the inspection. "It's no go," said Tom; "I tried
hard for it, too. Got up early in the morning—slapped twice round
the Park at a swinging pace. When I went before them it was red all
about, a couple of inches. The flinty-hearted villains gave me only a
gratuity, though it bled while they were looking at it."
At an early day after Pledget's and Tom's removal, we assembled
at the chateau, on an occasion in which we all felt a melancholy
interest—the funeral of Captain Gabion. The military arrangements,
of course, did not rest with us; Gingham had made every provision
which was left to his care with equal liberality and propriety.
Gingham also, no chaplain being present, officiated at the grave. He
read the service with great devoutness and solemnity. The
procession was joined, as we ascended the hill, by a mounted
officer, a major of the artillery, who, during the whole of the service,
seemed lost in thought, and stood with his eyes fixed upon the
coffin till it was lowered into the grave. The whole concluded, he
approached and shook hands with Gingham and myself, spoke a few
hurried words, took a hasty leave, mounted, and rode away.
Gingham and I waited by the grave till all was filled in and made
right; we then walked down together towards the city, both for some
time silent. I spoke first.
"Wouldn't it be right to communicate with the friends? I think they
ought to know the exact position of the grave, and also the
particulars which I got from my cousin."
"Why, yes," said Gingham; "it would, I think, be as well to give
them all the information you can. I have already written to the
widow."
 CHATEAUBRIAND'S MEMOIRS.
Mémoires d'outre Tombe. Par M. Le Vicomte de Chateaubriand. Tom.
v. vi. vii. viii. et ix. Paris: 1849.
The great and honourable feature of Chateaubriand's mind, amidst
some personal weaknesses, is its noble and disinterested character.
It differs from what we see around us, but it differs chiefly in
superior elevation. It united, to a degree which perhaps will never
again be witnessed, the lofty feelings of chivalry, with the
philanthropic visions of philosophy. In the tribune he was often a
Liberal of the modern school; but in action he was always a paladin
of the olden time. His fidelity was not to prosperity, but to adversity;
his bond was not to the powerful, but to the unfortunate; reversing
the revolutionary maxim, he brought the actions of public men to
the test, not of success, but of disaster. He often irritated his friends
when in power by the independence of his language, but he never
failed to command the respect of his enemies when in adversity, by
his constancy to misfortune. "Vive le roi quand-même," ever became
his principle when the gales of adversity blew, and the hollow-
hearted support of the world began to fail. Prosperity often saw him
intrepid, perhaps imprudent in expression, but misfortune never
failed to exhibit him generous and faithful in action; and his fidelity
to the cause of royalty was never so strikingly evinced as when that
cause in France was most desperate. He was the very antipodes of
the hideous revolutionary tergiversation of Fontainebleau. A pilgrim
in this scene of trial, he was ever ready, after having attained the
summit of worldly grandeur, to descend at the call of honour; and,
resuming his staff and scrip, to set out afresh on the path of duty.
He was fitted to be the object of jealousy and spite to kings and
ministers in power, whose follies he disdained to flatter or to
overlook their vices, and of eternal admiration to the great and the
good in every future age, whose hearts his deeds not less than his
words will cause to throb. Such a character might pass for fabulous
or imaginary, were it not clearly evinced, not only by words, but
actions; not only in the thoughts of genius, but in the deeds of
honour. His life, and the feelings by which it was regulated, are well
worth examining, although we fear he will find but few imitators in
these days, and is more likely, in a utilitarian and money-seeking
age, to be classed with the mammoth and mastodon, as a species of
existence never again to be seen in this world.
A character of this description naturally became enamoured of
awful or heartstirring events, and was ever ready to find a friend in
those capable of noble or heroic deeds in the ranks even of his
enemies. Both qualities are evinced in the following graphic account
of the appearance of the Grand Army when it arrived at Smolensko
during the Moscow retreat:—

"On the 9th November, the troops at length reached

Smolensko. An order of Buonaparte forbade any one to enter
before the posts had been intrusted to the Imperial Guard. The
soldiers on the outside were grouped in great numbers round
the foot of the walls: those within were under cover. The air
resounded with the imprecations of those who were shut out.
Clothed in dirty Cossack cloaks, horse-cloths, and worn-out
blankets, with their heads covered with old carpets, broken
helmets, ragged shakos, for the most part torn by shot, stained
with blood, or hacked in pieces by sabre-cuts—with haggard
and yet ferocious countenances, they looked up to the top of
the ramparts gnashing their teeth, with the expression of those
prisoners who, under Louis the Fat, bore in their right hand their
left cut off: you would have taken them for infuriated masques,
or famished madmen escaped from Bedlam. At length the Old
and Young Guard arrived, they were quickly admitted into the
place which had been wasted by conflagration on occasion of
our first passage. Loud cries of indignation were immediately
raised against the privileged corps. 'Is the army to be left
nothing but what it leaves?' was heard on all sides. Meanwhile
 the household troops, who had been admitted, rushed in
tumultuous crowds to the magazines like an insurrection of
spectres: the guards at the doors repulsed them; they fought in
the streets: the dead, the wounded encumbered the
pavements, the women, the children, the dying filled the
waggons. The air was poisoned by the multitude of dead
bodies; even old soldiers were seized with idiocy or madness;
some whose hair stood on end with horror, blasphemed, or
laughed with a ghastly air and fell dead. Napoleon let his wrath
exhale in imprecations against a miserable commissary, none of
the orders given to which had been executed.
"The army, a hundred thousand strong when it left Moscow,
now reduced to thirty thousand, was followed by a band of fifty
thousand stragglers; there were not eighteen hundred
horsemen mounted. Napoleon gave the command of them to M.
de Latour Maubourg. That officer, who had led the cuirassiers to
the assault of the great redoubt of Borodino, had had his head
almost cleft asunder by the stroke of a sabre; he afterwards lost
a leg at Dresden. Perceiving his servant in tears when the
operation was over, he said to him, 'Why do you weep? you will
have only one boot to clean.' That general, who remained
faithful to misfortune, became the preceptor of Henry V. in the
first years of the exile of that prince. I lift my hat in his
presence, as in that of the Incarnation of Honour."—Memoirs, vi.
p. 116, 118.

As Chateaubriand had declined office, and narrowly escaped death

in consequence, when Napoleon murdered the Duke d'Enghien, his
life, from that period to the Restoration of the Bourbons, was one of
retirement and observation. The important part which he took in the
Restoration, by the publication of his celebrated pamphlet De
Buonaparte et des Bourbons, restored him to political life. The effect
produced by that work was immense, and the placing of the ancient
race of monarchs on the throne was in a great degree owing to it;
for, at a crisis when the intentions of the Allies were yet undecided,
and Austria openly supported the strong party in France which
inclined for a regency with Marie Louise at its head, it swelled
immensely the numbers of the decided Royalists, and gave a definite
and tangible object to their hitherto vague and divided aspirations. It
was written with prodigious rapidity, and bears marks of the haste of
its composition in the vehemence of its ideas and the occasional
exaggeration of its assertions; but it was the very thing required for
a national crisis of unexampled importance, when every hour was
fraught with lasting consequences, and every effort of genius was
required for laying the foundation of a new order in European
society. Of the first conception and subsequent completion of this
remarkable work he gives the following account:—

"I had been permitted to return to my solitary valley. The

earth trembled under the footsteps of stranger armies: I wrote
like the last Roman, amidst the din of barbarian invasion. During
the day, I traced lines as agitated as the events which were
passing: at night, when the roar of cannon was no longer heard
in my solitary woods, I returned to the silence of the years
which sleep in the tomb, and to the peace of my earlier life. The
agitated pages which I wrote during the day, became, when put
together, my pamphlet On Buonaparte and the Bourbons. I had
so high an idea of the genius of Napoleon, and the valour of our
soldiers, that the idea of a foreign invasion, successful in its
ultimate results, never entered into my imagination; but I
thought that such an invasion, by making the French see the
dangers to which the ambition of Napoleon had exposed them,
would lead to an interior movement, and that the deliverance of
the French would be the work of their own hands. It was under
that impression that I wrote my notes, in order that, if our
political assemblies should arrest the march of the Allies, and
separate themselves from a great man who had become their
scourge, they should know to what haven to turn. The harbour
of refuge appeared to me to be in the ancient authority, under
which our ancestors had lived during eight centuries, but
modified according to the changes of time. During a tempest,
 when one finds himself at the gate of an old edifice, albeit in
ruins, he is glad to seek its shelter."—Vol. vi. p. 196, 197.

Madame de Chateaubriand, in a note, has described the

circumstances under which this memorable pamphlet was written,
and the morbid anxiety with which she was devoured during its
composition:—

"Had the pages of that pamphlet been seized by the police,

the result could not have been a moment doubtful: the
sentence was the scaffold. Nevertheless the author was
inconceivably negligent about concealing it. Often, when he
went out, he left the sheets on the table: at night he only
placed them under his pillow, which he did in presence of his
valet—an honest youth, it is true, but who might have betrayed
him. For my part, I was in mortal agonies: whenever M. de
Chateaubriand went out, I seized the manuscript, and concealed
it on my person. One day, in crossing the Tuileries, I perceived I
had it not upon me, and being sure I had it when I went out, I
did not doubt that I had let it fall on the road. Already I beheld
that fatal writing in the hands of the police, and M. de
Chateaubriand arrested. I fell down in swoon in the garden, and
some kind-hearted person carried me to my house, from which
I had only got a short distance. What agony I endured when,
ascending the stair, I floated between terror, which now
amounted almost to a certainty, and a slight hope that I might
have forgot the pamphlet. On reaching my husband's
apartment, I felt again ready to faint: I approached the bed—I
felt under the pillow; there was nothing there: I lifted the
mattress, and there was the roll of paper! My heart still beats
every time I think of it. Never in my life did I experience such a
moment of joy. With truth can I say, my joy would not have
been so great if I had been delivered at the foot of the scaffold,
for it was one who was more dear to me than life itself whom I
saw rescued from destruction."—Vol. vi. p. 206, 207.
 On the entrance of Louis XVIII. into Paris, on the 3d May 1814,
the Allied sovereigns, from a feeling of delicacy to that monarch,
gave orders that none but French troops should appear in the
procession. The Old Guard lined the streets next the palace, and
Chateaubriand gives the following account of the way in which they
received him:—

"A regiment of infantry of the Old Guard kept the ground,

from the Pont Neuf to Notre Dame, along the Quai des Orfures.
I do not believe that human figures ever expressed anything so
menacing and so terrible. These grenadiers, covered with
wounds, so long the terror of Europe, who had seen so many
thousand bullets fly over their heads, who seemed to smell of
fire and powder—these very men, deprived of their leader, were
forced to salute an old king, enfeebled by time and not
combats, guarded by an army of Russians, Austrians, and
Prussians, in the conquered capital of Napoleon! Some, shaking
their heads, made their huge bearskins fall down over their
eyes, so as not to see what was passing: others lowered the
extremities of their mouths, to express their contempt and rage:
others, through their mustaches, let their teeth be seen, which
they gnashed like tigers. When they presented arms, it was with
a gesture of fury, as if they brought them down to the charge.
The sound they made with the recover was like thunder. Never,
it must be admitted, had men been subjected to such a trial, or
suffered such a punishment. If, in that moment, they had been
called to vengeance, they would have exterminated the last
man, or perished in the attempt.
"At the extremity of the line was a young hussar on
horseback, with his drawn sabre in his hand; his whole body
literally quivered with a convulsive movement of wrath. He was
deadly pale; his eyes rolled round in the most frightful manner;
he opened his mouth alternately and shut it, grinding his teeth,
and uttering inarticulate cries of rage. He cast his eyes on a
Russian officer: no words can express the look which he gave
 him. When the carriage of the King passed before him, he made
his horse leap forward, it was easy to see that he withstood
with difficulty the temptation to precipitate himself on his
sovereign.[3]
"The Restoration, at its very outset, committed an irreparable
fault. It should have disbanded the army, preserving only the
marshals, generals, military governors, and officers, in their
rank, pay, and appointments. The soldiers, in this manner,
would have gradually re-entered their ranks, as they have since
done into the Royal Guard; but they would have done so
isolated from each other. The legitimate monarch would no
longer have had arrayed against him the soldiers of the empire
in regiments and brigades, as they had been during the days of
their glory, for ever talking to each other of times past, and
comparing the conquests of Napoleon with their inglorious
inactivity under their new master.
"The miserable attempt to reconstruct the Maison Rouge, that
mixture of the military men of the old monarchy and the
soldiers of the new empire, only augmented the evil. To
suppose that veterans famous on a hundred fields of battle
should not be shocked at seeing young men—brave without
doubt, but for the most part unaccustomed to the use of arms—
to see them wear, without having earned or deserved, the
marks of high military rank, was to be ignorant of the first
principles of human nature."—Vol. vi. p. 311-313.

These observations of Chateaubriand's are well founded, and the

last, in particular, is very important; but it may well be doubted
whether, by any measures that could have been adopted, the
support of the army could have been secured, or the dynasty of the
Bourbons established on a secure foundation. It was the fact of their
having been replaced by the bayonets of the stranger which was the
insurmountable difficulty; it was national subjugation, the capture of
Paris, which had for ever stained the white flag. This original sin in
its birth attended the Restoration through every subsequent year of
its existence: it was the main cause of the revolution of 1830, and
operated with equal force in bringing about the still more fatal one
of 1848. Impatience of repose—a desire to precipitate themselves on
foreign nations—an aversion to the employments and interests of
peace, were the secret but principal causes of these convulsions. If
either Louis XVIII. or Louis Philippe had been young and warlike
princes, and the recollection of Leipsic and Waterloo, of the
invasions of France, and the double capture of its capital, had not
prevented them from engaging in the career of foreign warfare; if
they had been enterprising and victorious, they would have secured
the unanimous suffrages of the nation, and continued the honoured
possessors of the throne of France. But this dazzling though perilous
career was denied to Louis XVIII. To him there was left only the
difficult, perhaps the impossible task, of reconciling irrevocable
enmities, of closing irremediable wounds, of appeasing
inextinguishable mortifications. They have been thus set forth in the
eloquent words of genius:—

"The house of Bourbon was placed in Paris at the Restoration

as a trophy of the European confederation. The return of the
ancient princes was inseparably associated, in the public mind,
with the cession of extensive provinces, with the payment of an
immense tribute, with the occupation of the kingdom by hostile
armies, with the emptiness of those niches in which the gods of
Athens and Rome had been the objects of a new idolatry, with
the nakedness of those walls on which the Transfiguration had
shone with a light as glorious as that which overhung Mount
Thabor. They came back to a land in which they could recognise
nothing. The Seven Sleepers of the legend, who closed their
eyes when the Pagans were persecuting the Christians, and
woke when the Christians were persecuting the Pagans, did not
find themselves in a world more completely new to them.
Twenty years had done the work of twenty generations. Events
had come thick; men had lived fast. The old institutions and the
old feelings had been torn up by the roots. There was a new
church founded and endowed by the usurper; a new nobility,
 whose titles were taken from the fields of battle, disastrous to
the ancient line; a new chivalry, whose crosses had been won
by exploits which seemed likely to make the banishment of the
Emigrants perpetual; a new code, administered by a new
magistracy; a new body of proprietors, holding the soil by a new
tenure; the most ancient local distinctions effaced, the most
familiar names obsolete. There was no longer a Normandy, a
Brittany, or a Guienne. The France of Louis XVI. had passed
away as completely as one of the Preadamite worlds. Its fossil
remains might now and then excite curiosity; but it was as
impossible to put life into the old institutions as to animate the
skeletons which are imbedded in the depths of primeval strata.
The revolution in the laws and the form of government was but
an outward sign of that mightier revolution which had taken
place in the minds and hearts of men, and which affected every
transaction and feeling of life. It was as absurd to think that
France could again be placed under the feudal system, as that
our globe could be overrun by mammoths. Louis might efface
the initials of the Emperor, but he could not turn his eyes
without seeing some object which reminded him he was a
stranger in the palace of his fathers."[4]

As a parallel to this splendid passage, though in an entirely

different style, we gladly give place to a noble burst of
Chateaubriand, on that most marvellous of marvellous events, the
return of Napoleon from Elba. It was natural that so memorable a
revolution should strongly impress his imaginative mind; but he
seems to have exceeded himself in the reflections to which it gives
rise. We know not whether to award the prize to the Englishman or
the Frenchman, in these parallel passages. They are both
masterpieces in their way. Perhaps the correct view is, that Macaulay
is superior in graphic force and the accumulation of sarcastic
images; Chateaubriand in lofty thought and imaginative images.

"On the 1st March, at three o'clock in the morning, Napoleon

approached the coast of France in the Gulf of Juan; he
 disembarked, walked along the shore, gathered a few violets,
and bivouacked in an olive wood. The inhabitants withdrew in a
state of stupefaction. He left Antibes to his left, and threw
himself into the Mountains of Grasse in Dauphiny. At Sisterone
the road passes a defile where twenty men might have stopped
him; he did not meet a living soul. He advanced without
opposition among the inhabitants who the year before had
wished to murder him. Into the void which was formed around
his gigantic shade, if a few soldiers entered, they straightway
yielded to the attraction of his eagles. His fascinated enemies
seek him and find him not; he shrowds himself in his glory, as
the lion in the Sahara desert conceals himself in the rays of the
sun to dazzle the eyes of his pursuers. Enveloped in a burning
halo, the bloody phantoms of Arcola, Marengo, Austerlitz, Jena,
Friedland, Eylau, the Moskwa, Lützen, and Bautzen, form his
cortege amidst a million of the dead. From the midst of that
column of smoke and flame, issue at the gates of towns some
trumpet-notes mingled with tricolor standards, and the gates fly
open. When Napoleon passed the Niemen, at the head of four
hundred thousand foot, and a hundred thousand horse, to blow
into the air the palace of the Czars at Moscow, he was less
wonderful than when, breaking his ban, casting his fetters as a
gauntlet in the face of kings, he came alone from Cannes to
Paris, to sleep peaceably in the palace of the Tuileries."—Vol. vi.
p. 359, 360.

To a mind like that of Chateaubriand, reposing in solitude when

Napoleon was acting with such marvellous effect in the world, the
character and qualities of that wonderful man could not fail to be a
constant object of solicitude and observation. It has been already
noticed that he braved the Emperor in the plenitude of his power,
and essentially contributed, in the crisis of his fate, to his
dethronement, and the re-establishment of the ancient line of
princes. But, as is not unusual with persons of his highly wrought
and generous temper of mind, his hostility to the Emperor declined
with the termination of his authority, and his admiration for his
genius rose with the base desertion of the revolutionary crowd who
had fawned upon him when on the throne. The following
observations on the style of his writings, indicate the growth of this
counter feeling, and are in themselves equally just and felicitous:—

"His partisans have sought to make of Buonaparte a perfect

being; a model of sentiment, of delicacy, of morality, and of
justice—a writer like Cæsar and Thucydides, an orator like
Demosthenes, a historian like Tacitus. The public discourses of
Napoleon, his sonorous phrases in the tent and at the council
board, are the less inspired by the spirit of prophecy, that many
of the catastrophes which he announced have not been
accomplished, while the warlike Isaiah himself has disappeared.
Prophecies of doom which follow without reaching states
become ridiculous. It is their accomplishment which renders
them sublime. During sixteen years, Napoleon was the
incarnation of destiny. Destiny now is mute, and he, too, should
be so. Buonaparte was not a Cæsar; his education had neither
been learnedly nor carefully conducted: half a stranger, he was
ignorant of the first rules of our language, and could hardly spell
it; but what did it signify, after all, that his expression was
defective?—he gave the law to the universe. His bulletins have
the most thrilling of all eloquence—that of victory. Sometimes,
during the intoxications of success, they affected to be written
on a drum-head: in the midst of the most lugubrious accents,
something emerged which excites a smile. I have read all that
Napoleon has written—the first manuscripts of his infancy, his
love-letters to Josephine, the five volumes of his discourses,
bulletins, and orders; but I have found nothing which so truly
portrays the character of that great man, when in adversity, as
the following autograph note left at Elba:—
"'My heart refuses to share in ordinary joys as ordinary
sorrows.
"'Not having given myself life, I am not entitled to take it
away.
 "'My bad genius appeared to me and announced my end;
which I found at Leipsic.
"'I have conjured up the terrible spirit of innovation, which will
overrun the world.'
"Certes, there is Napoleon to the very life. His bulletins and
discourses have often great energy; but it was not his own; it
belonged to the age; he only adopted it. It sprang from the
revolutionary energy, which he only weakened by moving in
opposition to it. Danton said, 'The metal is fused; if you do not
watch over the furnace, you will be consumed.' St Just replied,
'Do it if you dare.' These words contain the whole secret of our
Revolution. Those who make revolutions by halves, do nothing
but dig their own graves."—Vol. vii. p. 101.

Certes, there is Chateaubriand to the very life.

Chateaubriand, as all the world knows, was Minister for Foreign
Affairs to Louis XVIII. at Ghent; adhering thus to his ruling maxim
throughout life, "Fidelity to misfortune." So great were the services
rendered by him to the cause of European freedom, by the energetic
series of papers which he poured forth with unwearied vigour every
week, that there were serious thoughts, after the battle of Waterloo,
of promoting him to the dignity of Prime Minister. Louis XVIII. openly
inclined to it; and if his advice had prevailed, the catastrophe which
fifteen years afterwards befel his family, would probably have been
prevented. But the insuperable difficulty lay here: the pure and
honourable mind of Chateaubriand revolted from the idea of forming
a Ministry in conjunction with Talleyrand and Fouché; and yet their
influence was such that the monarch, in the first instance at least,
was compelled to court their assistance. Expedience, at least
immediate expedience, seemed to counsel it; but Chateaubriand,
animated by higher principles, and gifted with a more prophetic
mind, anticipated no lasting advantage, but rather the reverse, from
an alliance with the arch-regicide of Nantes, and the arch-traitor who
had sworn allegiance to and betrayed twelve Governments in
succession. But the chorus of "base unanimities," as he expresses it,
with which the monarch was surrounded, proved too strong for any
single individual, how gifted soever. Fouché and Talleyrand were
taken into power, and Chateaubriand retired. Of the conversation
with Louis XVIII., when this vital change was resolved on, he gives
the following interesting account, which proves that that sagacious
monarch at least was well aware of the consequences of the step to
which he was thus involuntarily impelled:—
"Before quitting St Denis, on our way back to Paris, I had an
audience of the King, and the following conversation ensued:

"'Well?' said Louis XVIII., opening the dialogue by that

exclamation.
"'Well, sire, you have taken the Duke of Otranto,' (Fouché.)
"'I could not avoid it; from my brother to the bailie of Crussol,
(and he at least is not suspected,) all said that we could not do
otherwise—what think you?'
"'Sire! the thing is done; I crave permission to remain silent.'
"'No, no—speak out; you know how I resisted at Ghent.'
"'In that case, sire, I must obey my orders. Pardon my
fidelity: I think it is all over with the monarchy.'
"The King remained some time silent. I began to tremble at
my boldness, when his Majesty rejoined:—
"'In truth, M. de Chateaubriand, I am of your opinion.'
"I bowed and withdrew; and thus ended my connection with
the Hundred Days."—Vol. vii. 70.

Manzoni has written an ode, known over all Europe, on the double
fall of Napoleon: "The last poet," says Chateaubriand, "of the
country of Virgil, sang the last warrior of the country of Cæsar.
 Tutte ei provo, la gloria
Maggior dopo il periglio,
La fuga e la Vittoria,
La reggia e il triste esiglio:
Due volte nella polvere,
Due volte sugli altar.

Ei se nomo: due secoli,

L'un contro l'altro armato,
Sommessi a lui se volsero,
Come aspettando il fato:
Ei fe silenzio ed arbitro
S'assise in mezzo a loro.

"He proved everything; glory greater after danger, flight, and

victory: Royalty and sad exile, twice in the dust, twice on the
altar.
"He announced himself: two ages, armed against each other,
turned towards him, as if awaiting their fate; he proclaimed
silence, and seated himself as arbiter between them."

Notwithstanding the vehemence of Chateaubriand's dissension

with Napoleon, it cannot be expected that a man of his romantic and
generous temperament would continue his hostility after death. No
one, accordingly, has awarded a more heartfelt or magnanimous
tribute to his memory.

"The solitude of the exile and of the tomb of Napoleon has

shed an extraordinary interest, a sort of prestige, over his
memory. Alexander did not die under the eyes of Greece, he
disappeared amidst the distant wonders of Babylon. Buonaparte
has not died under the eyes of France: he has been lost in the
gloomy edge of the southern horizon. The grandeur of the
silence which now surrounds him equals the immensity of the
noise which his exploits formerly made. The nations are absent:
 the crowd of men has retired: the bird of the tropics,
"harnessed," in Buffon's words, "to the chariot of the sun," has
precipitated itself from the star of light—where does it now
repose? It rests on the ashes of which the weight has all but
subverted the globe."

"Imposuerunt omnes sibi diademata post mortem ejus; et

multiplicata sunt mala in terrâ."[5] "They all assumed diadems after
his death, and evils were multiplied on the earth." Twenty years
have hardly elapsed since the death of Napoleon, and already the
French and Spanish monarchies are no more. The map of the world
has undergone a change: a new geography is required: severed
from their legitimate rulers, nations have been thrown against
nations: renowned actors on the scene have given place to ignoble
successors: eagles from the summits of the loftiest pines have
plunged into the ocean, while frail shellfish have attached
themselves to the sides of the trunk, which still stands erect.

"As in the last result everything advances to its end, 'the

terrible spirit of innovation which overruns the world', as the
Emperor said, and to which he had opposed the barrier of his
genius, has resumed its course. The institutions of the
conqueror fail: he will be the last of great existences on the
earth. Nothing hereafter will overshadow society, parcelled out
and levelled: the shadow of Napoleon alone will be seen on the
verge of the old world which has been destroyed, like the
phantom of the deluge on the edge of its abyss. Distant
posterity will discern that spectre through the gloom of passing
events still erect above the gulf into which unknown ages have
fallen, until the day marked out by Providence for the
resurrection of social man."—Vol. vii. 169-171.

Assuredly no one can say that Chateaubriand's genius has

declined with his advanced years.
To a man viewing Napoleon with the feelings expressed in these
eloquent words, the translation of his remains from their solitary
resting-place under the willow at St Helena could not but be an
object of regret. He thus expresses himself on that memorable
event, and future ages will probably confirm his opinion:—

"The removal of the remains of Napoleon from St Helena was

a fault against his renown. A place of sepulchre in Paris can
never equal the Valley of Slanes. Who would wish to see the
Pillar of Pompey elsewhere than above the grave dug for his
remains by his poor freedman, aided by the old legionary? What
shall we do with those magnificent remains in the midst of our
miseries? Can the hardest granite typify the everlasting duration
of Napoleon's renown? Even if we possessed a Michael Angelo
to design the statue on the grave, how should we fashion the
mausoleum? Monuments are for little men, for the great a stone
and a name. At least they should have suspended the coffin
from the summit of the triumphal arch which records his
exploits: nations from afar should have beheld their master
borne aloft on the shoulders of his victories. Was not the urn
which contained the ashes of Trajan placed at Rome, beneath
his column? Napoleon at Paris will be lost amidst the crowd of
unknown names. God forbid he should be exposed to the
vicissitudes of our political changes, surrounded though he is by
Louis XIV., Vauban, and Turenne. Let a certain section of our
revolutionists triumph, and the ashes of the conqueror will be
sent to join the ashes which our passions have dispersed. The
conqueror will be forgotten in the oppressor of our liberties. The
bones of Napoleon will not reproduce his genius; they will only
teach his despotism to ignoble soldiers."—Vol. vii. 184, 185.

The Restoration did not immediately employ Chateaubriand. His

anticipations were realised. The chorus of baseness and selfishness
with which the court was surrounded, kept him at a distance. They
were afraid of his genius: they were jealous of his reputation. Above
all, they dreaded his independence. He was not sufficiently
manageable. They were actuated, perhaps not altogether without
reason, by the same feeling which made Lord North say, when urged
to bring Dr Johnson into Parliament, whose great powers in the
political warfare of pamphlets had been so signally evinced on the
side of Government, "No, sir, he is an elephant, but a wild one, as
likely to trample under foot his friends as his enemies." The veteran
statesman, so well versed in the ways of men, was right. Genius is
the fountain of thought: it ultimately rules the councils and destinies
of men; but it generally requires to be tempered by time before it
can be safely introduced into practice.
Chateaubriand enlivens this period of his memoirs, which is
neither signalised by political event nor remarkable literary effort, by
a sort of biography of Madame Recamier, with whom he was on
terms of intimate friendship. This remarkable person, who was
beyond all question the most beautiful and attractive woman of her
age in France, or perhaps in Europe, is now no more; and he
appears to have obtained from her relatives, or perhaps from herself
prior to her decease, not only many curious and highly interesting
details concerning her early years and subsequent history, but a
great variety of original letters from the most eminent men of the
age, who were successively led captive by her charms, but none of
whom appear to have impaired her reputation. In this country,
where the lines of severance between the sexes are much more
rigidly drawn, it would be impossible for a young and beautiful
married woman to be in the habit of receiving the most ardent love-
letters from a great variety of distinguished and fascinating admirers,
without the jealousy of rivals being excited, and the breath of
scandal fastening upon her as its natural prey. But it is otherwise on
the Continent, where, although there is doubtless abundance of
dissoluteness of manners in certain circles, yet in others such
intimacies may exist, which are yet kept within due bounds, and cast
no reflection on the fortunate fair one who sees all the world at her
feet.
Such, at least, appears to have been the case with Madame
Recamier, the intimate friend of Madame de Stael, who said "She
would willingly give all her talents for one half of her beauty;" and
whose powers of fascination were such, that she not only inspired a
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com