Datasets:
PV_Multi_Defect-corrected: 1044 images / 3403 boxes / 5 classes. Removed 61 byte-identical duplicates and 3 orphan/empty files, kept the near-duplicate cluster on one side of a seeded 70/15/15 stratified split, cleaned VOC metadata (personal hard-coded path + bogus folder), added YOLO TXT + classes.txt + data.yaml + CSVs
4d5eac6 | # Ultralytics YOLO dataset configuration for PV_Multi_Defect-corrected | |
| # Cleaned and standardized by KeenForge (https://github.com/KeenForgeAI) | |
| path: . | |
| train: images/train | |
| val: images/val | |
| test: images/test | |
| nc: 5 | |
| names: ['black_border', 'broken', 'hot_spot', 'no_electricity', 'scratch'] | |