Datasets:
TrafficSigns_YOLO-corrected: 6000 images / 6000 boxes / 6 classes. Replaced the upstream split, which scattered near-identical Roboflow augmentations across train/valid/test, with a leak-safe seeded 71/14/15 stratified split (4259/858/883). Added Pascal VOC XML + classes.txt + data.yaml + CSVs.
346cf09 | # Ultralytics YOLO dataset configuration for TrafficSigns_YOLO-corrected | |
| # Cleaned and standardized by KeenForge (https://github.com/KeenForgeAI) | |
| path: . | |
| train: images/train | |
| val: images/val | |
| test: images/test | |
| nc: 6 | |
| names: ['left_turn', 'straight_ahead', 'right_turn', 'dead_end', 'no_entry', 'stop'] | |