openai / CLIP

CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image

shap / shap

A game theoretic approach to explain the output of any machine learning model.