bardisk/gloria

GLoRIA: A Multimodal Global-Local Representation Learning Framework forLabel-efficient Medical Image Recognition

A Project For The Evalution of GLoRIA

This projects are used for evalute GLoRIA in various datasets.

Plug in datasets

Refer to datasets/README.md.

We prepared an example dataset(vindr, of a size about 100M) for you here at NJUBOX. You can directly extract the tarball to ./datasets/.

Download the pretrained weights

The original respository of GLoRIA provides the download link the pretrained weights of GloRIA on chexpert dataset.

You should put the files in ./pretrained/.

Prepare the environment

conda env create -f environment.yml
conda activate gloria_test
pip install -r requirements.txt -i https://mirror.nju.edu.cn/pypi/web/simple

Execute the evaluation

make

The results for all datasets plugged will be output to ./results/.