Apppublic

Thiratpy/hemoscan-inference-api

sourceHugging Faceupdated 3mo agoView on Hugging Face

0likes

App README

HemoScan Inference API

FastAPI service used by the HemoScan Next.js app for blood-smear image analysis:

—Anemia classification
—RBC morphology (9-class)
—Leukemia (C-NMC ALL) detection — soft-voting ensemble of 3 CNNs (EfficientNetV2-S, ConvNeXt-Tiny, DenseNet121). Binary detection only (Hem/normal vs ALL/malignant) — not a leukemia subtype classifier. Research / educational demo, not a medical device.

Endpoints

—GET /health — status + loaded model info
—POST /predict — body {"image": "<data-url base64>"} → combined prediction JSON

Local dev

bash

pip install -r requirements.txt
uvicorn main:app --host 0.0.0.0 --port 7860 --reload

This Space is built from the python-api/ directory of the HemoScan repo. The 3 leukemia ensemble weight files are tracked with Git LFS (see .gitattributes).