Computer Vision Project - DLBAIPCV01

This is the code repository to support the project report for module: DLBAIPCV01 – Project: Computer Vision at I.U. International University of Applied Sciences.

Overview

In this project, we utilize various object detection models, including YOLOv8, Faster R-CNN, and Single Shot Detector, to perform predictions on custom datasets and videos.

Repository Contents

Scripts

• object_detector.py: Contains the code to make predictions on an input video using the YOLOv8 model.
• torchvision_models_evaluator.py: Contains the code to make predictions on a dataset using torchvision's pretrained models.
• yolo_evaluator.py: A simple script to evaluate the YOLOv8 model on a custom dataset. The evaluation.yaml file points to the custom dataset.
• calculate_metrics.py: A simple script which utilizes the pycocotools library to compute the key metrics used for evaluation.

Media Files

• desk.mp4: An example video used for prediction.
• output_video.mp4: The result video with the annotated frames.

Folders

• evaluation/: Holds the custom dataset, formatted according to YOLO standards.
• evaluation_results/: Contains the results of evaluation for the three models (YOLOv8, Faster R-CNN, Single Shot Detector).
• resources/: Contains the ground truth of the COCO 2017 validation dataset as a JSON file. A modified version with unnecessary items removed (e.g., licenses, info) is also present.

Getting Started

Prerequisites

• Python 3.x
• PyTorch
• torchvision
• YOLOv8 (ultralytics package)
• pycocotools

Dataset

The evaluation/ folder contains the custom dataset formatted according to YOLO standards. Ensure that your dataset follows the same structure if you plan to use your own data.

Evaluation Results

The evaluation_results/ folder contains the results of evaluation for the three models:

• YOLOv8
• Faster R-CNN
• Single Shot Detector

Resources

The resources/ folder contains:

• The ground truth of the COCO 2017 validation dataset (instances_val2017.json).
• A modified version with unnecessary items removed (e.g., licenses, info).

Acknowledgments

• I.U. International University of Applied Sciences
• PyTorch
• Ultralytics YOLOv8