search

duyhominhnguyen / LVM-Med

[NeurIPS 2023] Release LMV-Med pre-trained models
188 stars 24 forks source link

Using the resnet-50 model #6

Closed gowtham07 closed 10 months ago

gowtham07 commented 1 year ago

Hello,

Thank you so much for the pre-trained models on Medical dataset. I would like to use the resnet-50, vit models to finetune on a downstream task like [Brain Tumor Classification (MRI)] etc . My aim is to generate sailency features from different models to see if they kind of learn same thing. For this it is important that the model did not see in pre-training. I would like to know if model was pretrained on these dataset (Kvasir, BUID, FGADR, MMWHS_MR_Heart and MMWHS_CT_Heart, brain tumor classification) .

Thanks!!

duyhominhnguyen commented 1 year ago

Hi @gowtham07, sorry for the late response.

You can check the Table in the Appendix where we summarized all of dataset used in the pre-training stage. Generally, there are a few datasets seen in the pre-training stage, like MMWHS_MR, MMWHS-CT, etc; however, we excluded samples used in a pre-training stage in the fine-tuning step; therefore, data leaking should not occur! Let me know if you have further questions. Thank you.