YOLO-NAS object detection on custom dataset

This project is centered around a YOLO-NAS (You Only Look One - Neural Architecture Search) object detection model. It offers a comprehensive guide for building and training a custom YOLO-NAS model tailored to detect specific objects. In this instance, the model is trained to recognize street signs and classify them into four categories.

DEMO

Training and Validation

Utilize the traffic.ipynb notebook for training and validating the custom model. This notebook, originally crafted by AarohiSingla for fall detection, requires Python 3.9 or higher.

Inference

Once trained, the model is ready for use with the flask_app. The flask_app folder contains app.py, allowing users to upload images or videos for object detection. Configure settings in the config folder.

References and Datasets

1. Original Code built for fall detection: https://github.com/AarohiSingla
2. Dataset for traffic sign detection: https://www.kaggle.com/datasets/valentynsichkar/traffic-signs-dataset-in-yolo-format
3. YOLO-NAS: https://learnopencv.com/yolo-nas/
4. Video demo: https://www.youtube.com/watch?v=40xZVEFVBuE