Augmentations for Mammo-CLIP

batmanlab / Mammo-CLIP

Official Pytorch implementation of MICCAI 2024 paper (early accept, top 11%) Mammo-CLIP: A Vision Language Foundation Model to Enhance Data Efficiency and Robustness in Mammography

https://shantanu-ai.github.io/projects/MICCAI-2024-Mammo-CLIP/

Creative Commons Attribution 4.0 International

35 stars 11 forks source link

Augmentations for Mammo-CLIP #14

Closed Smith6791 closed 2 months ago

Smith6791 commented 2 months ago

Hi, Great work. Can u please specify the specific augmentations u use for pre-training/classification/detection? Also i read this paper: Ladder (https://arxiv.org/pdf/2408.07832). Do you use the same Mammo-CLIP to probe the failure modes in ladder as well? Or there is any difference? Bcz there is no code for Ladder repo. I believe you are the same author for both the papers.

shantanu-ai commented 2 months ago

Hi @Smith6791

for preraining, follow this
for classification/detection, follow this

Also, for detection, use the CSV file specified in the readme as we have resized the images, so we reshaped the bounding boxes accordingly.

For your query on ladder, yes we are the same group who created ladder. We are doing more comprehensive evaluations for ladder, so the draft in arxiv will be edited soon. Ladder uses Mammo-CLIP to probe the classifier to discover the error slices. Also keep an eye of the repo of ladder as the code will also be uploaded soon there.

Smith6791 commented 2 months ago

Hi, Thanks for reply. Can u plz give me the folder structure of the datasets u used?

shantanu-ai commented 2 months ago

Hi @Smith6791 Look into this