MedSigLIP Demo

Zero-shot classification and finetuning for medical images using MedSigLIP.

Frank Li | HITI Lab, Emory University | Datathon 2025

Usage

cd ~
git clone https://github.com/f10409/MedSigLIP_demo.git

Launch Jupyter and select the medsiglip kernel to run the demo.

If the pre-installed environment doesn't work:

cd ~/MedSigLIP_demo
conda deactivate
uv sync
source .venv/bin/activate
python -m ipykernel install --user --name medsiglip

Launch Jupyter and select the medsiglip kernel to run the demo.

Name		Name	Last commit message	Last commit date
Latest commit History 16 Commits
models		models
notebook_figs		notebook_figs
utils		utils
weights		weights
.gitignore		.gitignore
0_0_hugging_face.ipynb		0_0_hugging_face.ipynb
0_1_wandb.ipynb		0_1_wandb.ipynb
1_zero_shot_classification.ipynb		1_zero_shot_classification.ipynb
2_extract_image_embeddings.ipynb		2_extract_image_embeddings.ipynb
3_fine-tuning.ipynb		3_fine-tuning.ipynb
4_finetuning_evaluation.ipynb		4_finetuning_evaluation.ipynb
LICENSE		LICENSE
README.md		README.md
__init__.py		__init__.py
keys_.py		keys_.py
pyproject.toml		pyproject.toml
requirements_medsiglip.txt		requirements_medsiglip.txt
uv.lock		uv.lock