CSC173-DeepCV-Montemayor

[Project Title: e.g., Real-Time Object Detection for Waste Sorting]

CSC173 Intelligent Systems Final Project
Mindanao State University - Iligan Institute of Technology
Student: [Your Full Name], [Student ID]
Semester: [e.g., AY 2025-2026 Sem 1]

Abstract

[150-250 words: Summarize problem (e.g., “Urban waste sorting in Mindanao”), dataset, deep CV method (e.g., YOLOv8 fine-tuned on custom trash images), key results (e.g., 92% mAP), and contributions.][web:25][web:41]

Table of Contents

• Introduction
• Related Work
• Methodology
• Experiments & Results
• Discussion
• Ethical Considerations
• Conclusion
• Installation
• References

Introduction

Problem Statement

[Describe the real-world vision problem and why it matters locally (e.g., IoT waste systems in Iligan).]

Objectives

• [Objective 1: e.g., Achieve >90% detection accuracy]
• [Objective 2: Integrate with decision logic]
• [Objective 3: Deploy on edge device]

[web:41]

Related Work

• [Paper 1: YOLOv8 for real-time detection [1]]
• [Paper 2: Transfer learning on custom datasets [2]]
• [Gap: Your unique approach, e.g., Mindanao-specific waste classes] [web:25]

Methodology

Dataset

• Source: [e.g., Custom 5K images + COCO subset]
• Split: 70/15/15 train/val/test
• Preprocessing: Augmentation, resizing to 640x640 [web:41]

Architecture

• Backbone: [e.g., CSPDarknet53]
• Head: [e.g., YOLO detection layers]
• Hyperparameters: Table below

Parameter	Value
Batch Size	16
Learning Rate	0.01
Epochs	100
Optimizer	SGD

Training Code Snippet

train.py excerpt model = YOLO(‘yolov8n.pt’) model.train(data=’dataset.yaml’, epochs=100, imgsz=640)

Experiments & Results

Metrics

| Model | mAP@0.5 | Precision | Recall | Inference Time (ms) | |——-|———|———–|——–|———————| | Baseline (YOLOv8n) | 85% | 0.87 | 0.82 | 12 | | Ours (Fine-tuned) | 92% | 0.94 | 0.89 | 15 |

Demo

[Video: CSC173_YourLastName_Final.mp4] [web:41]

Discussion

• Strengths: [e.g., Handles occluded trash well]
• Limitations: [e.g., Low-light performance]
• Insights: [e.g., Data augmentation boosted +7% mAP] [web:25]

Ethical Considerations

• Bias: Dataset skewed toward plastic/metal; rural waste underrepresented
• Privacy: No faces in training data
• Misuse: Potential for surveillance if repurposed [web:41]

Conclusion

[Key achievements and 2-3 future directions, e.g., Deploy to Raspberry Pi for IoT.]

Installation

1. Clone repo: git clone https://github.com/yourusername/CSC173-DeepCV-YourLastName
2. Install deps: pip install -r requirements.txt
3. Download weights: See models/ or run download_weights.sh [web:22][web:25]

requirements.txt: torch>=2.0 ultralytics opencv-python albumentations

References

[1] Jocher, G., et al. “YOLOv8,” Ultralytics, 2023.
[2] Deng, J., et al. “ImageNet: A large-scale hierarchical image database,” CVPR, 2009. [web:25]

GitHub Pages

View this project site: https://jjmmontemayor.github.io/CSC173-DeepCV-Montemayor/ [web:32]