dpc_im — Mouse Skin H&E Segmentation Models

Fine-tuned models for semi-automated segmentation of mouse skin layers and hair follicles from H&E histology images. Part of the dpc_im pipeline.

Models

File	Task	Base model	Training images
best_model.pth	5-class layer segmentation (background, dermis, epidermis, adipose, muscle)	U-Net ResNet34 — ImageNet	5 annotated mouse skin H&E images (1920×1440 px, 4×, 7.615 µm/px)
cellpose_hf_best	Hair follicle instance segmentation	Cellpose SAM	8 train + 2 val annotated mouse skin H&E images (1920×1440 px, 4×, 7.615 µm/px)

Usage

See dpc_im for installation and full pipeline usage.

Citation

Pachitariu, M., Rariden, M., & Stringer, C. (2025). Cellpose-SAM: superhuman generalization for cellular segmentation. bioRxiv. https://doi.org/10.1101/2025.04.03.647135

Iakubovskii, P. (2019). Segmentation Models Pytorch. GitHub. https://github.com/qubvel/segmentation_models.pytorch