Dataset Format Converter & Leak-Free Splitter
Convert object-detection datasets between YOLO, COCO and Pascal VOC in one command, and split any image dataset into train/val/test with stratification and duplicate grouping, so near-identical images never leak between training and test sets. Outputs a ready-to-train data.yaml or COCO JSON. Runs locally, never modifies your source files.
#Techniek#Onderzoek#Analyse
Beoordeling
Meer beoordelingen nodig
Verkocht
0
Gebruikswijze
Downloaden
Dataset Format Converter & Leak-Free Splitter
Get your dataset into the format and split your training framework expects, in one command, without near-duplicate images leaking between train and test.
What it does
- Convert object-detection annotations between YOLO, COCO and Pascal VOC (boxes are converted and validated; out-of-range boxes are clipped or reported).
- Split any detection or classification dataset into train / val / test with class stratification.
- Leak-free grouping: exact and near-duplicate images (perceptual hash) are always kept in the same split, so your test score is honest.
- Stats: images per class, boxes per image, box sizes, empty images.
- Writes a ready-to-train data.yaml (Ultralytics YOLO) or COCO instances JSON (Detectron2 / MMDetection).
Safe by design
- Read-only on your source folder; output always goes to a new folder.
- Runs 100% locally. Your images never leave your machine.
Example requests
- "Convert my VOC dataset to YOLO and make an 80/10/10 split."
- "My val accuracy looks too good. Re-split without duplicates leaking."
- "Turn this COCO JSON into YOLO labels for Ultralytics."
Requirements
Python 3.8+, numpy, Pillow. Replies in English, Traditional/Simplified Chinese, Japanese, Korean, Vietnamese or Indonesian.
Pairs well with Dataset Label Audit to find mislabels before training.
