Datasets:
CPLID_Insulators-corrected: 845 images / 1566 boxes / 2 classes (insulator, defect). Removed 3 pixel-identical duplicates, grouped consecutive UAV-capture frames so bursts stay on one side of the split, merged the paired defect+insulator label folders into one annotation per image, added a seeded 70/15/15 stratified split, cleaned VOC metadata (bogus folder + hard-coded personal path), removed a stray 48-byte file
955a16c | # Ultralytics YOLO dataset configuration for CPLID_Insulators-corrected | |
| # Cleaned and standardized by KeenForge (https://github.com/KeenForgeAI) | |
| path: . | |
| train: images/train | |
| val: images/val | |
| test: images/test | |
| nc: 2 | |
| names: ['defect', 'insulator'] | |