Cell Imaging Lab
Segment cells and nuclei in microscopy images with Cellpose-SAM. Upload an image, get instance masks, bounding boxes, and per-object statistics — through one API key, no CUDA setup.
What's inside
Cellpose-SAM Segmentation
Cell and nucleus segmentation for any microscopy image. Returns instance masks, bounding boxes, areas, and per-object mean intensity.
Multiple Model Variants
cyto3 (general), nuclei (DAPI/Hoechst), cyto2 (legacy), livecell (time-lapse), tissuenet (histology) — all exposed via a single diameter + model parameter.
Batch-Ready
Upload images as base64, receive JSON with per-object stats. No CUDA setup, no container management, no weight downloads.
Built for Microscopy Postdocs
The microscopy crowd is the largest underserved scientific audience on the open web. Cell Imaging Lab is the first API-first aggregator that serves them.
FAQ
What input format do I need?›
Base64-encoded PNG or JPEG microscopy image. Pass it in the image_b64 field of the POST /v1/imaging/cellpose request. Max 10 MB per image.
Which Cellpose model should I use?›
'cyto3' is the default and works for general cell segmentation in brightfield or fluorescence. Use 'nuclei' for DAPI/Hoechst-stained nuclei, 'livecell' for time-lapse, 'tissuenet' for histology. 'cyto2' is the legacy model for backward compatibility.
What's the 'diameter' parameter?›
Approximate cell/nucleus diameter in pixels. Setting this correctly dramatically improves segmentation quality. Default is 30 px. If unsure, try a few values and see which gives the best object count.
Is my image data kept?›
No. Images are processed in-memory and not persisted. Only the segmentation output (object count + bounding boxes + stats) is returned. Nothing is logged.
Can I get the actual segmentation masks?›
Sprint 51 returns bounding boxes + per-object stats. Full per-pixel instance masks via a paginated output endpoint are coming in Sprint 51B. For now, bounding boxes are enough for most downstream analysis (cell counting, density, area distribution).