Nipple segmentation (furry)

A segmentation model trained with yolo11m with a dataset of furry images.

Dataset contains different female, male, colored, shaped, sized and also pierced nipples. With some different view angles too, so sideboob should work, though likely not as good.

Trained on an initial dataset of ~450 images (training: 666(augmented), validation: 220, test: 15) for 90 epochs.

mAP50(Box): 0.892

mAP50-95(Box): 0.679

mAP50(Mask): 0.878

mAP50-95(Mask): 0.619