Slides presented in the All Japan Computer Vision Study Group on May 15, 2022. Methods for disentangling the relationship between multimodal data are discussed.
Slides presented in the All Japan Computer Vision Study Group on May 15, 2022. Methods for disentangling the relationship between multimodal data are discussed.
【DLゼミ】XFeat: Accelerated Features for Lightweight Image Matchingharmonylab
公開URL:https://arxiv.org/pdf/2404.19174
出典:Guilherme Potje, Felipe Cadar, Andre Araujo, Renato Martins, Erickson R. ascimento: XFeat: Accelerated Features for Lightweight Image Matching, Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)
概要:リソース効率に優れた特徴点マッチングのための軽量なアーキテクチャ「XFeat(Accelerated Features)」を提案します。手法は、局所的な特徴点の検出、抽出、マッチングのための畳み込みニューラルネットワークの基本的な設計を再検討します。特に、リソースが限られたデバイス向けに迅速かつ堅牢なアルゴリズムが必要とされるため、解像度を可能な限り高く保ちながら、ネットワークのチャネル数を制限します。さらに、スパース下でのマッチングを選択できる設計となっており、ナビゲーションやARなどのアプリケーションに適しています。XFeatは、高速かつ同等以上の精度を実現し、一般的なラップトップのCPU上でリアルタイムで動作します。
GPU の分析への応用などの基礎技術の進化とクラウドの爆発的な普及に伴い、だれもが使いたいときに使いたい時だけ高性能なマシンリソースを使える時代が到来し、家電、スマホ、ビジネスアプリケーションなどありとあらゆるものに AI が搭載されているとうたわれ、一部のデータサイエンティストが担っていた高度な分析や深層学習のフレームワークもエンドユーザーで使いこなす人も少なくありません。
一方で、AI や深層学習という言葉が独り歩きし、まず AI 導入ありきでプロジェクトが始まり、目的が失われ頓挫するようなケースや、予測した結果についての妥当性について説明がつかず、結果がうまく利用できないようなケースも見られるようになってきました。
今回のセミナーでは、AI や高度な分析についての最新トレンドと、その使いどころについて、実際の事例や経験などを踏まえお伝えします。
This document summarizes a paper titled "DeepI2P: Image-to-Point Cloud Registration via Deep Classification". The paper proposes a method for estimating the camera pose within a point cloud map using a deep learning model. The model first classifies whether points in the point cloud fall within the camera's frustum or image grid. It then performs pose optimization to estimate the camera pose by minimizing the projection error of inlier points onto the image. The method achieves more accurate camera pose estimation compared to existing techniques based on feature matching or depth estimation. It provides a new approach for camera localization using point cloud maps without requiring cross-modal feature learning.
2020/10/10に開催された第4回全日本コンピュータビジョン勉強会「人に関する認識・理解論文読み会」発表資料です。
以下の2本を読みました
Harmonious Attention Network for Person Re-identification. (CVPR2018)
Weekly Supervised Person Re-Identification (CVPR2019)
2018/10/20コンピュータビジョン勉強会@関東「ECCV読み会2018」発表資料
Yew, Z. J., & Lee, G. H. (2018). 3DFeat-Net: Weakly Supervised Local 3D Features for Point Cloud Registration. European Conference on Computer Vision.
5. Voxelベースの手法 (1/2)
[Maturana2015]Maturana, D., & Scherer, S. (2015).VoxNet:
A 3D Covolutional Neural Network for Real-Time Object
Recognition. In International Conference on Intelligent Robots
and Systems.
[Li2017]Li, B. (2017). 3D fully convolutional network for
vehicle detection in point cloud. IEEE International
Conference on Intelligent Robots and Systems
[Zeng2015]Zeng Wang, D., & Posner, I. (2015).Voting for
Voting in Online Point Cloud Object Detection. Robotics:
Science and Systems XI.
6. Voxelベースの手法 (2/2)
[Engelcke2017]Engelcke, M., Rao, D.,Wang, D. Z.,Tong, C.
H., & Posner, I. (2017).Vote3Deep: Fast object detection in
3D point clouds using efficient convolutional neural
networks. IEEE International Conference on Robotics and
Automation, (September),
[Zhou2018]Zhou,Y., & Tuzel, O. (2018).VoxelNet: End-to-
End Learning for Point Cloud Based 3D Object Detection.
In Conference on ComputerVision and Pattern Recognition.
[Yan2018]Yan,Y., Mao,Y., & Li, B. (2018). SECOND: Sparsely
Embedded Convolutional Detection. Sensors, 18(10)
8. [Zeng2015]Voting for Voting (1/2)
入力点群(+反射率)を
Voxel化し、3D Sliding
Windowで物体検出
各VoxelごとにHand-Crafted
特徴量(Grid内に点が存在
するか、反射率平均、反射
率分散、3種のShape Factor*
の計6種)を算出し、Sliding
Window内でそれらを結合し、
線形SVMで判別
N個の向きに対して演算
入力点群 Voxel化
Voxel特徴ベクトル
3D Sliding Window
*C.-F.Westin, S. Peled, H. Gudbjartsson, R. Kikinis, and F.
A. Jolesz,“Geometrical Diffusion Measures for MRI
fromTensor Basis Analysis,” in ISMRM ’97,Vancouver
Canada,April 1997, p. 1742.
9. [Zeng2015]Voting for Voting (2/2)
SlidingWindow + 線形SVMは畳み込み演算とみなせ、入
力が疎な場合、投票で高速処理
a. 赤、緑、水色の個所にのみ点群が存在する場合、Window
のアンカー(青)上のスコアはこれらの重み付き線形和であ
らわされる
b. データの存在する個所(赤)は青位置のアンカーに投票する
17. Bird’s Eye Viewベースの手法
[Yang2018]Yang, B., Luo,W., & Urtasun, R. (2018). PIXOR: Real-time
3D Object Detection from Point Clouds. In IEEE conference on
ComputerVision and Pattern Recognition
[Luo2018]Luo,W.,Yang, B., & Urtasun, R. (2018). Fast and Furious: Real
Time End-to-End 3D Detection,Tracking and Motion Forecasting
with a Single Convolutional Net. In Conference on ComputerVision
and Pattern Recognition.
[Ren2018]Ren, M., Pokrovsky,A.,Yang, B., & Urtasun, R. (2018). SBNet:
Sparse Blocks Network for Fast Inference. In IEEE Conference on
ComputerVision and Pattern Recognition (pp. 8711–8720).
[Yang2018_2]Yang, B., Liang, M., & Urtasun, R. (2018). HDNET :
Exploiting HD Maps for 3D Object Detection. In Conference on Robot
Learning (pp. 1–10).
[Simon2018]Simon, M., Milz, S.,Amende, K., & Gross, H. (2018).
Complex-YOLO:An Euler-Region-Proposal for Real-time 3D Object
Detection on Point Clouds.ArXiv, arXiv:1803.
26. その他の手法
[Li2016]Li, B., Zhang,T., & Xia,T. (2016).Vehicle Detection
from 3D Lidar Using Fully Convolutional Network.
Robotics Science and Systems.
[Kunisada2018]Kunisada,Y.,Yamashita,T., & Fujiyoshi, H.
(2018). Pedestrian-Detection Method based on 1D-CNN
during LiDAR Rotation. In International Conference on
IntelligentTransportation Systems (ITSC).
45. 紹介しきれなかった研究(1/3)
1. Spinello, L.,Arras, K. O.,Triebel, R., & Siegward, R. (2010).A Layered
Approach to People Detection in 3D Range Data. In AAAI
Conference on Artificial Intelligence (pp. 1635-1630).
2. Teichman,A., & Thrun, S. (2011).Tracking-based semi-supervised
learning. In Robotics: Science and Systems.
3. Teichman,A., Levinson, J., & Thrun, S. (2011).Towards 3D object
recognition via classification of arbitrary object tracks. Proceedings
- IEEE International Conference on Robotics and Automation,
4034-4041.
4. Wang, D. Z., Posner, I., & Newman, P. (2012).What could move?
Finding cars, pedestrians and bicyclists in 3D laser data. Proceedings
- IEEE International Conference on Robotics and Automation,
4038-4044.
5. Behley, J., Steinhage,V., & Cremers,A. B. (2013). Laser-based Segment
Classification Using a Mixture of Bag-of-Words. In International
Conference on Intelligent Robots and Systems.
46. 紹介しきれなかった研究(2/3)
6. Asvadi,A., Garrote, L., Premebida, C., Peixoto, P., & Nunes, U. J.
(2017). DepthCN :Vehicle Detection Using 3D-LIDAR and
ConvNet. In International Conference on IntelligentTransportation
Systems (ITSC).
7. Zidan, M. I., & Sallab,A.A.Al. (2018).YOLO3D : End-to-end real-time
3D Oriented Object Bounding Box Detection Object Bounding
Box Detection from LiDAR, (August).
8. Feng, D., Rosenbaum, L.,Timm, F., & Dietmayer, K. (2018). Leveraging
Heteroscedastic Aleatoric Uncertainties for Robust Real-Time
LiDAR 3D Object Detection.ArXiv, arXiv:1809.
9. Yun, P.,Tai, L.,Wang,Y., & Liu, M. (2018). Focal Loss in 3D Object
Detection.ArXiv, arXiv:1809.
10. Feng, D., Rosenbaum, L., & Dietmayer, K. (2018).Towards Safe
Autonomous Driving: Capture Uncertainty in the Deep Neural
Network For Lidar 3DVehicle Detection. International Conference
on IntelligentTransportation Systems (ITSC).
47. 紹介しきれなかった研究(3/3)
11. Minemura, K., Liau, H., Monrroy,A., & Kato, S. (2018). LMNet : Real-
time Multiclass Object Detection on CPU using 3D LiDAR. In 3rd
Asia-Pacific Conference on Intelligent Robot Systems (ACIRS).
12. Gustafsson, F., & Linder-Norén, E. (2018). Automotive 3D Object
DetectionWithoutTarget Domain Annotations. Linköping University.
13. Zeng,Y., Hu,Y., Liu, S.,Ye, J., Han,Y., Li, X., & Sun, N. (2018). RT3D:
Real-Time 3DVehicle Detection in LiDAR Point Cloud for
Autonomous Driving. IEEE Robotics and Automation Letters, 3766(c),
14. Beltr, J., Guindel, C., Moreno, F. M., Cruzado, D., Garc, F., & Escalera,
A. De. (2018). BirdNet : a 3D Object Detection Framework from
LiDAR information. ArXiv, arXiv:1805.
15. Wirges, S., Fischer,T., & Stiller, C. (2018). Object Detection and
Classification in Occupancy Grid Maps using Deep Convolutional
Networks. ArXiv, arXiv:1805.