Human Segmentation Tool

Upload an image and select a task. The model detects and segments human figures using a YOLO26 + UNET pipeline, fine-tuned on a combined dataset of LIP, COCO-person, MADS, and Penn-Fudan images.

Note: Running on CPU — inference takes 3-10 seconds per image. Results are best on images with clearly visible, upright people.

Select Task

Outputs a binary black/white mask. White pixels = person, black = background. Useful for creating ground truth masks for datasets.

Example Images

Model: YOLO26s (detection) + UNET ResNet-50 (segmentation)
Training data: LIP · COCO-person · MADS · Penn-Fudan
GitHub: Human Segmentation Evaluation