This project builds a highly robust object detection system designed to perform under challenging real-world conditions such as fog, snow, glare, nighttime, and occlusion. It is inspired by the perception challenges in autonomous vehicles like Waymo and Zoox.
configs/: Training configs (YAML/JSON).data/: Data loaders, augmentation transforms, and utilities.models/: Wrapper or custom model code.train/: Training scripts and loss function definitions.evaluate/: Evaluation metrics and visualization tools.scripts/: Data downloading, formatting, and conversion utilities.notebooks/: EDA and experiment notebooks.outputs/: Saved checkpoints, logs, and result images.
git clone <your-repo-url>
cd robust_object_detection
pip install -r requirements.txt
python main.py --mode train --config configs/default_config.yaml