Vehicle Detection in the Context of Bangladesh

GitHub Repository

Models used: YOLOv6, YOLOv8, Co-DETR, RTDETR

• Developed a pipeline for detecting vehicles in the context of Bangladesh using the BadODD dataset, which contains over 5,000 images captured across 9 cities in the country.
• Experimented with several pretrained models, including YOLO variants and detection transformers. RTDETR was selected as the most suitable model, as it effectively balances the strengths of both YOLO models and DETRs.