COCO-format noisy-label benchmarks for instance segmentation (VIPER-N / COCO-N)

[mkimhi]

Hi! I’m sharing a small HF collection of datasets/benchmarks for instance segmentation under noisy labels (COCO-format):

• Collection: https://huggingface.co/collections/Kimhi/noisy-labels-for-instance-segmentation-coco-format

It contains:

• VIPER (clean): images + clean COCO instances_*.json
• VIPER-N: noisy COCO instance annotations (annotations-only)
• COCO-N: noisy COCO instance annotations (annotations-only)

All annotations are standard COCO instances JSON, so you can evaluate by swapping instances_{train,val}2017.json paths.

If this is relevant to your project, I’d love any feedback—and if you maintain a benchmark list / eval suite, feel free to include the collection.

Thanks!