Autonomous Traffic Data Detection and Tracking: YOLO NAS, DEEP SORT, and SODA 10M.

Figure 1: Input Video

Figure 2: Output Video

Overview

This repository contains the implementation of 2D object detection and tracking on traffic data, specifically using the SODA 10M dataset. We demonstrate fine-tuning the YOLO NAS model for object detection and applying the Deep SORT algorithm for 2D tracking of traffic data.

Dataset

The dataset used is the SODA 10M dataset, which can be found at SODA 2D official website. It is designed for Autonomous traffic data analysis.

Repository Structure

.
├── checkpoints/yolonas_AI # Checkpoints for the YOLO NAS model
├── data # Data used for training and testing models
├── deep_sort_pytorch # Deep SORT algorithm implementation
├── .gitignore # Specifies intentionally untracked files to ignore
├── README.md # The README file you are currently reading
├── data_preprocessing.ipynb # Notbook for data preprocessing, converting from COCO to YOLO format.
├── detection_NAS_utils.py # Utility functions for YOLO NAS detection
├── detection_tracking_from_vedio.ipynb # Notebook for detection and tracking from video
├── nas_sort_utils.py # Utility functions for NAS and SORT integration
├── test_finetune_result.ipynb # Notebook for testing fine-tuning results
├── train_NAS_utils.py # Utility functions for training YOLO NAS
├── train_nas.ipynb # Notebook for training YOLO NAS
└── video_predictions.yaml # Output predictions in YAML format

Features

• Fine-tuning of YOLO NAS model for autonomous traffic data detection.
• 2D tracking using the Deep SORT algorithm.
• Comprehensive Jupyter notebooks that guide through the detection and tracking process.
• Output predictions stored in a YAML file for further analysis.

Getting Started

To get started with this project, clone this repository using:

git clone https://github.com/MautushiD/AI_project.git

Prerequisites

python 3.10+

numpy

pandas

scikit-learn

matplotlib

tensorflow

torch

super-gradients

ultralytics

pathlib

yaml

cv2

torchvision

Installation

Install the required dependencies:

pip install -r requirements.txt

Follow the instructions in each notebook to train the models and perform detection and tracking.

Train the YOLO NAS model:

Run the train_nas.ipynb notebook to train the model on the SODA 10M dataset. Detection and Tracking: Use detection_tracking_from_vedio.ipynb to perform detection and tracking on video data. The output will be saved in ./test_video_output and ./video_predictions.yaml. Acknowledgements This project utilizes the SODA 10M dataset, obtained from the SODA 2D official website. We thank the creators for making the dataset publicly available for research purposes.

Contributing

Contributions to this project are welcome. Please submit a pull request or create an issue for any features or improvements.