YOLOv8 Car Object Detection

This repository provides scripts for training and evaluating YOLOv8 models on a car-object detection dataset. The project is built using the Ultralytics YOLOv8 library and integrates with WandB for experiment tracking.

Setup

Installation

1. Clone the repository:

   git clone https://github.com/your-username/yolov8-car-object-detection.git
   cd yolov8-car-object-detection

2. Install dependencies:

   pip install -r requirements.txt

3. Download the car-object-detection dataset on google drive and extract:

   Car Object Detetion Dataset

   This dataset downloaded from Kaggle and have formatted into YOLO format.

4. Configure your WandB API key:

   Open config.conf and replace YOUR WANDB KEY with your actual WandB API key. Replace {/path/to/data}/car-object-detection with data path downloaded above.

Training

To train the YOLOv8 model, run:

python trainer.py

Adjust the configuration in config.conf to suit your needs.

Evaluation

To evaluate the trained model and visualize predictions, run:

python evaluate.py --yolo_model runs/detect/train/weights/best.pt --testing_paths "runs/detect/predict"

Replace yolov8m.pt with the desired trained model weights file.

Directory Structure

• data: Directory for storing the car-object-detection dataset.
• runs: Directory for storing training and evaluation results.

Evaluate Result

Acknowledgments

• Ultralytics YOLO
• WandB
• Car Object Detection Dataset