Car damage detection and segmentation with YOLO on the CarDD dataset: eight YOLO11/YOLO26 models benchmarked against the paper (best 58.4 mask AP), pretrained weights included, car-part detection for panel names, and a live Streamlit demo that turns a photo of a damaged car into a damage report.
benchmark machine-learning computer-vision deep-learning pytorch yolo object-detection automotive pretrained-models image-segmentation instance-segmentation onnx insurance-claims streamlit car-damage-detection ultralytics yolo11 yolo26 vehicle-damage-assessment cardd
-
Updated
Nov 11, 2026 - Python