FYDP (Phase-I) Progress Report [Version 1.

0]

DAFFODIL INTERNATIONAL UNIVERSITY

FYDP (Phase-I) Progress Report
REPORTING PERIOD- SPRING 2025

Project Identification:
I. Project Title Deep Learning-Based Detection of Recyclable Plastics
using Image with YOLOv8
II. Group Members 1. Name: Tawhidul Islam Student ID:213-15-4311
2. Name: Robiullah Roni Student ID:213-15-4329
III. Supervisor Name: Mr. Md. Mezbaul Islam Zion
Designation: Lecturer
IV. Co-Supervisor Name: Dr. Md. Taimur Ahad
Designation: Associate Professor
V. Submission Date:
VI. Certificate : “This is to certify that the final year design project work
until Phase-I evaluation held on 24th May 2025, titled as
stated in Sec. I, executed by the students’ group
(Signature of Supervisor
mentioned in Sec. II, have been found satisfactory and & date)
every section of this report is reflecting the same.”

Project Insights
Thematic Area(s): Artificial Intelligence and Machine Learning ☒
[Just click the check box]
Data Science and Analytics ☐
Cybersecurity ☐
Software Engineering and Development ☐
Blockchain Technology ☐
Internet of Things (IoT) ☐
Computer Networks ☐
Human-Computer Interaction (HCI) ☐
Big Data Technologies ☐
Computer Vision ☐
Natural Language Processing (NLP) ☐
Robotics ☐
Game Development ☐
Cloud Computing ☐
Biomedical Computing ☒

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 1
FYDP (Phase-I) Progress Report [Version 1.0]

Others :
1. Smart Waste Management
2. Automation and Intelligent Systems

Software packages, Programming Language: Python

tools, and Software and Platforms:
programming 1. Google Colab
languages 2. Roboflow
3. Google Drive
Packages List:
1. Ultralytics(YOLOv8)
2. OpenCV(sv2)
3. NumPy
4. Pandas
5. PyYAML
6. torch

CO Description for FYDP-Phase-I

CO CO Descriptions PO
CO1 Integrate recently gained and previously acquired knowledge to identify a PO1
Deep Learning Approach for Recyclable Plastic Detection problem for the
Final Year Design Project (FYDP)
CO2 Analyze different aspects of the goals in designing a solution for the FYDP PO2
CO3 Explore diverse problem domains through a literature review delineate the PO4
issue and establish the goals for the FYDP
CO4 Perform economic evaluation and cost estimation and employ suitable project PO11
management procedures throughout the development life cycle of the FYDP

1. Project Overview:
1.1 Introduction

Plastic pollution continues to pose a serious threat to environmental and public health,
especially in developing countries where recycling infrastructure is limited. Conventional
waste sorting systems often rely on manual labor or costly sensor-based mechanisms like
infrared spectrometry and barcode readers, which are impractical for widespread use due to
their expense and complexity. In light of these challenges, the integration of deep learning
and computer vision into waste management has emerged as a promising alternative.

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 2
FYDP (Phase-I) Progress Report [Version 1.0]

Recent advancements in object detection models—particularly the YOLO (You Only Look
Once) family—have shown impressive results in real-time detection tasks. YOLOv8, the
latest in the series, offers improved speed, accuracy, and model compactness, making it
suitable for real-time plastic classification using image inputs from everyday devices such as
smartphones.

Unlike prior systems that depend on industrial-grade setups like Reverse Vending Machines
(RVMs), this study explores a lightweight framework for plastic waste detection tailored for
low-cost deployment. The aim is to enable small-scale users, including local recycling
centers and municipal units, to automate plastic sorting efficiently. By focusing on image-
based classification alone, the system bypasses the financial and operational barriers
associated with hardware-dependent solutions. This research contributes to the sustainable
development goals by promoting practical, AI-powered waste management accessible to
low-income communities.

1.2 Background

Bangladesh produces over 800,000 tons of plastic waste annually, yet less than 30% is
effectively recycled. Manual sorting methods are inefficient and error-prone. Several studies
have proposed methods for automating recyclable waste classification using a mix of
convolutional neural networks and YOLO-based object detection. Kestane et al. [1]
implemented YOLOv8 in Reverse Vending Machines (RVMs), training on a combined
TrashNet and custom dataset; their system achieved 97.5% accuracy, outperforming
YOLOv5, YOLOv7, and YOLOv9 in the same task. Kwon and Kim [2] developed a smart
bin using a webcam on a Raspberry Pi running YOLO, attaining 91.0% accuracy under
optimal conditions (75.0% when fully deployed on-device).

Mwangi and Mokoena [3] applied a lightweight SqueezeNet CNN on Raspberry Pi–captured
images of PET bottles, achieving 98.0% status recognition accuracy. Kunwar et al. [4]
trained a YOLO-11m model on the WaDaBa dataset, reporting 98.03% overall accuracy and
an mAP@50 of 0.990. Larbi et al. [5] compared various CNN architectures against YOLOv5
for plastic waste detection, finding CNNs reached up to 95.0% accuracy but suffered from
slower inference times.

Riyadi et al. [6] benchmarked YOLOv5 versus YOLOv8 on recyclable waste classification
and found YOLOv8 to consistently outperform its predecessor in both detection speed and
accuracy. Zhang et al. [7] (ResNet18 + Spatial Mixed Module) demonstrated 95.87%
accuracy on the TrashNet dataset for multi-class waste classification. Zhang S. et al. [8]
applied VGG16 to household waste sorting under a four-category regulation, achieving
94.7% accuracy. Kim et al. [9] designed a dual-image classification system combining
SqueezeNet and MobileNetV2, reaching 95.0% accuracy and highlighting the benefits of
multi-view inputs.

These works collectively illustrate the progression from sensor-dependent and single-CNN
methods toward real-time, YOLO-based solutions. Building on these insights, our study
deploys YOLOv8 for recyclable plastic detection using only standard image inputs from
widely available devices, thereby maximizing accessibility and minimizing cost.

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 3
FYDP (Phase-I) Progress Report [Version 1.0]

2. Objectives:
i. Develop a YOLOv8-based recyclable plastic detection model using image input
only.
ii. Build a custom dataset with labeled plastic types found in Bangladesh.
iii. Evaluate the model's performance against established metrics (accuracy, precision,
recall).
iv. Explore real-world deployment potential using a smartphone or web interface.

Methodology/ Requirement Specification:

2.1 Research Design/ Prototype Design
Adopt a computer vision–based deep learning pipeline using YOLOv8. Prepare a
labeled image dataset and train the model on local plastic categories.
Data Collection/ Need Assessment
Images were collected using a smartphone and digital camera. Categories include PET
bottles, shampoo bottles (HDPE/PP), plastic snack wrappers (multi-layered), etc. Data
annotated using Roboflow tools.
2.2 Analysis Techniques
Training and validation using YOLOv8 via Ultralytics framework. Metrics: mAP@50,
precision, recall, inference speed. Model comparisons with prior YOLO versions (v5, v7)
using benchmark tests.

Progress Achieved:
2.3 Completed Tasks
 Defined research scope and objectives.
 Reviewed 17 relevant research papers [1] – [17].
 Collected and annotated over 100 images of plastic waste.
 Prepared initial training environment using Google Colab and YOLOv8.

2.4 Results Obtained

Verified feasibility of image-only detection using basic cameras. Identified optimal
conditions for dataset collection (lighting, angles, backgrounds).

Challenges Faced:
S.No. Issues and Challenges Strategies or Plans
1 Lighting variation and glare in images Use side lighting, avoid backlighting
2 Lack of labeled datasets specific to Bangladesh Build custom dataset with local
examples

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 4
 FYDP (Phase-I) Progress Report [Version 1.0]

3 Resource limitation (no GPU PC) Use Google Colab for training
4 Similar-looking plastic types (e.g., PET vs PP) Use diverse samples and
augmentations
3. Next Steps:
- Outline the tasks and milestones planned for the next phase of the project.
S.No. Next Task Estimate completion time
(MM-YY)
1 Complete image collection (min. 500 samples) 06-2025
2 Finalize annotation and labeling 06-2025
3 Train and optimize YOLOv8 model 07-2025

4 Build basic web app or mobile interface 08-2025

4. Updated Timeline:

Task Week
1 2 3 4 5 6 7 8 9 12 13 14 15 16 17 18
Planning &
Research
Data Collection

Data Prepocessing

Model Setup &

Training
Testing

Documentation

Final Submission

Estimated Work Period

Actual Work Period

5. Resources Utilized:
 Smartphone with 13MP camera

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 5
FYDP (Phase-I) Progress Report [Version 1.0]

 Digital camera
 Roboflow for labeling
 Google Colab (YOLOv8 + Ultralytics)
 Python, OpenCV, and Ultralytics libraries

6. Project Management and Financial Analysis:

Estimated cost: ~BDT 0 (tools used are free/open-source). Potential cost includes data
storage or cloud GPU use. Project managed via Google Sheets.

7. Future Considerations:
Possible inclusion of more plastic types. Expansion to mixed waste classification.
Deployment optimization for mobile devices and integration with recycling infrastructure.
And also, deployment of Reverse Vending Machine (RVM)

8. Conclusion:
The project is progressing according to plan. Initial results confirm that a YOLOv8-based
model using images alone can detect recyclable plastics effectively. The upcoming phase will
focus on dataset expansion and model optimization.

References
[1] B. B. Kestane, E. Guney, and C. Bayilmis, “Real-time Recyclable Waste Detection Using
YOLOv8 for Reverse Vending Machines,” Jurnal Ilmiah Teknik Elektro Komputer dan
Informatika, vol. 10, no. 2, pp. 345–358, 2024.
https://journal.uad.ac.id/index.php/JITEKI/article/view/29208
[2] H. Kwon and S. Kim, “YOLO-Based Object Detection for Separate Collection of Recyclables
and Capacity Monitoring of Trash Bins,” Electronics, vol. 11, no. 9, p. 1323, 2024.
https://www.mdpi.com/2079-9292/11/9/1323
[3] H. W. Mwangi and M. Mokoena, “Using Deep Learning to Detect Polyethylene
Terephthalate (PET) Bottle Status for Recycling,” Global Journal of Computer Science and
Technology, vol. 19, no. 1, pp. 23–30, 2019.
https://computerresearch.org/index.php/computer/article/view/1890
[4] S. Kunwar, B. R. Owabumoye, and A. S. Alade, “Plastic Waste Classification Using Deep
Learning: Insights from the WaDaBa Dataset,” arXiv preprint arXiv:2412.20232, 2024.
https://arxiv.org/abs/2412.20232
[5] G. Larbi, G. A. Aridj, and L. Saher, “CNN-based Plastic Waste Detection System,” BOHR
International Journal of Smart Computing and Information Technology, vol. 3, no. 1, pp. 15–22,
2022. https://bohrpub.com/journals/BIJSCIT/2022/paper3

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 6
FYDP (Phase-I) Progress Report [Version 1.0]

[6] S. Riyadi, A. D. Andriyani, and A. M. Masyhur, “Classification of Recyclable Waste Using

Deep Learning: A Comparison of YOLO Models,” Revue d'Intelligence Artificielle, vol. 38, no.
4, pp. 569–576, 2024. https://iieta.org/journals/ria/paper/10.18280/ria.380404
[7] Q. Zhang et al., “A Waste Image Classification Model Based on ResNet18 and SMM,”
Resources, Conservation & Recycling, vol. 171, 2021.
doi:10.1016/j.resconrec.2021.105636.https://doi.org/10.107816/j.resconrec.2021.105636
[8] S. Zhang et al., “Computer Vision-Based Classification and Automatic Sorting of Household
Waste Using VGG16,” Resources, Conservation & Recycling, vol.
169,2021.doi:10.1016/j.resconrec.2021.105543.https://doi.org/10.1016/j.resconrec.2021.105543
[9] H. Kim et al., “Dual Image-Based Waste Detection Using SqueezeNet and MobileNetV2,”
Applied Sciences, vol. 11, no. 22, 2021. doi:10.3390/app112211051.
https://doi.org/10.3390/app112211051
[10] K. Erin, B. Bingöl, and B. Boru, “YOLO-Based Waste Detection,” Journal of Smart
Systems Research, vol. 3, no. 2, pp. 45–53,
2022.https://www.researchgate.net/publication/366595849_YOLO_-Based_Waste_Detection

[11] P. K. Yadav et al., “Plastic Contaminant Detection in Aerial Imagery of Cotton Fields with
Deep Learning,” arXiv preprint, arXiv:2212.07527, 2023.https://arxiv.org/abs/2212.07527

[12] E. Z. Kuang, K. R. Bhandari, and J. Gao, “Optimizing Waste Management with Advanced
Object Detection for Garbage Classification,” arXiv preprint, arXiv:2410.09975,
2024.https://arxiv.org/abs/2410.09975

[13] H. Rasool, A. Wahid, and M. U. Ghafoor, “Application of Computer Vision Techniques for
Segregation of Plastic Waste based on Resin Identification Code,” arXiv preprint,
arXiv:2011.07747, 2020.https://arxiv.org/abs/2011.07747

[14] X. Zhao et al., “An Intelligent Waste-Sorting and Recycling Device Based on Improved
EfficientNet,” Sensors, vol. 22, no. 20, 7974, 2022.https://www.mdpi.com/1424-8220/22/20/7974

[15] J. Chen et al., “A Vision Detection Scheme Based on Deep Learning in a Waste Plastics
Sorting System,” Applied Sciences, vol. 13, no. 7, 2023. https://www.mdpi.com/2076-
3417/13/7/4634
[16] M. Prakash et al., “Enhancing Waste Sorting and Recycling Efficiency: Robust Deep
Learning-Based Approach for Classification and Detection,” Neural Computing and

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 7
FYDP (Phase-I) Progress Report [Version 1.0]

Applications, 2024.
https://link.springer.com/article/10.1007/s00521-024-10855-2

[17] R. T. Kaur and A. Singh, “Application of Machine Learning in Plastic Waste Detection and
Classification: A Systematic Review,” Processes, vol. 12, no. 8, 1632,
2024.https://www.mdpi.com/2227-9717/12/8/1632

Appendix
- [Additional documentation, charts, graphs, or technical details]

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 8
FYDP (Phase-I) Progress Report [Version 1.0]

FINAL YEAR DESIGN PROJECT

PHASE-I PROGRESS REPORT
This report, in the form of a template, has been specifically designed for BSc. students working
on their Final Year Design Project (FYDP) at Computer Science and Engineering Department,
Daffodil International University (DIU).

Every group of students is required to do the following:

1. Complete all the sections of this template
2. Get it certified by the assigned supervisor before one week of Phase-I evaluation
presentations
3. Submit 01 photocopy to each of the following, on or before the day of Phase-I
presentations:
a. Supervisor
b. Internal Evaluator
4. Submit original copy to FYDP committee on the day of Phase-I presentations.

Note:
1. Use English
2. There should be NO grammatical or spelling mistakes
3. Submission after due date will not be accepted
4. For more information, contact your Supervisor

Template prepared by: Template approved by:

FYDP Committee Dr. Sheak Rashed Haider Noori

Dept. of CSE, DIU Professor and Head, Dept. of CSE, DIU

The students and faculty members of Computer Science and Engineering Department, Daffodil International University have full
access rights to read and print this document without any prior notice to the Head and FYDP committee.

All rights are reserved.

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 9
FYDP (Phase-I) Progress Report [Version 1.0]

© Copyright 2024 Department of Computer Science and Engineering, Daffodil International University 10