YOLO Models for Autonomous Driving

This project is a comprehensive toolkit for researchers and developers working on autonomous driving datasets. It automates the conversion of different file formats into the YOLO format and facilitates training of various YOLO model versions.

Features

• Multi-format Converter:
  Convert common autonomous driving dataset annotations (e.g., Pascal VOC, COCO, KITTI, BDD100K into the YOLO format.

• YOLO Trainer:
  Seamlessly train different versions of YOLO (v10, v11, ·v12, etc.) using pre-configured training pipelines.

• Modular and Extensible:
  Designed with modularity in mind, allowing you to add new conversion modules or support additional YOLO versions as needed.

Getting Started

Prerequisites

• Python 3.10
• Git
• uv

Installation

git clone [email protected]:vs-uulm/yolo-for-autonomous-driving.git
cd yolo-for-autonomous-driving
uv sync

Usage Example

Download and Convert KITTI Dataset to YOLO Format

uv run convert_dataset_to_yolo_format.py ./config/converter/kitti.yaml

Train YOLOv11s with KITTI from Scratch and Pretrained from COCO

uv run train_yolos.py ./config/trainer/yolov11_kitti.yaml

Check The Model Quality

cd runs