CellMap - HHMI Janelia

The CellMap project team at HHMI Janelia Research Campus develops machine learning models for segmentation of cellular structures in volume electron microscopy (vEM) data.

Models

We provide pre-trained deep learning models for 3D organelle segmentation, available in multiple formats:

Available Segmentation Targets

• Mitochondria
• Endoplasmic Reticulum (ER)
• Nucleus
• Peroxisomes
• Plasma Membrane
• Vesicles
• And more...

Quick Start

pip install cellmap-models

from cellmap_models.model_export.cellmap_model import CellmapModel
from huggingface_hub import snapshot_download

# Download and load any model
path = snapshot_download(repo_id="cellmap/your-model-name")
model = CellmapModel(path)

# Inference
output = model.ts_model(input_tensor)

# Finetuning
trainable_model = model.train()

Resources

• cellmap-models (GitHub) - Model architectures, weights, and export tools
• CellMap Project Page
• OpenOrganelle - Browse our EM datasets

Publications

• Heinrich, L. et al. Whole-cell organelle segmentation in volume electron microscopy. Nature (2021)

Contact

For questions or collaborations, visit our GitHub or reach out through the Janelia CellMap project page.