Datasets:
DBA_Fire_Smoke-corrected: 3846 images / 7988 boxes / 2 classes (flame 5211, smoke 2777). Removed 59 byte-identical duplicates, grouped 3014 near-duplicate clusters so video frames / re-encodes stay on one side of a seeded 70/15/15 stratified split, normalised 521 filenames containing spaces/parentheses, added Pascal VOC XML + classes.txt + data.yaml + CSVs
0d6fe62 | # Ultralytics YOLO dataset configuration for DBA_Fire_Smoke-corrected | |
| # Cleaned and standardized by KeenForge (https://github.com/KeenForgeAI) | |
| path: . | |
| train: images/train | |
| val: images/val | |
| test: images/test | |
| nc: 2 | |
| names: ['flame', 'smoke'] | |