UCell

A small cell segmentation model with not so small generalizability.

Ref: https://doi.org/10.48550/arXiv.2604.00243

Try it

docker run --gpus=all -p 50051:50051 jiyuuchc/ucell:latest

This creates an GRPC endpoint at port 50051, which you can visit using GUI clients, e.g. napari-biopb.

If you prefer programatic interface: see proto document here and code example here.

Fine-tuning

We recommend this setting

python train.py \
  --init ${BASE_MODEL} \
  --config config.py:train \
  --config.data_dir=${DATADIR} \
  --config.n_iters=1 \
  --config.epochs_per_iter=1024 \
  --config.ema_decay=0.95 \
  --config.lora.rank=16

Save your training data (*.tif, *_label.tif) under ${DATADIR}/train.

Based model weights can be downloaded here.

---