Steel surface defect detection using YOLOv8m on the Severstal dataset. Full pipeline: RLE-to-bbox conversion, image tiling, stratified splitting, custom balanced sampling, and Gradio deployment on HuggingFace Spaces.
python opencv computer-vision deep-learning end-to-end pytorch kaggle deployed object-detection huggingface dataset-analysis datasets-preparation severstal ultralytics steel-defect-detection full-pipeline huggingface-spaces yolov8
-
Updated
Jun 7, 2026 - Jupyter Notebook