search

cvlab-stonybrook / PromptMIL

Repository for "Prompt-MIL: Boosting Multi-Instance Learning Schemes via Task-specific Prompt Tuning" (MICCAI2023)
16 stars 0 forks source link

packages #3

Closed Rokinluohhh closed 6 months ago

Rokinluohhh commented 6 months ago

Can you provide the version of the packages(such as timm)? Looking forward to your reply!

jingweizhang-xyz commented 6 months ago

We provided the version of pytorch-lightning related packages (timm==0.6.12), and the rest does not matter. If you encounter any problems, please let me know.

Rokinluohhh commented 6 months ago

Thanks for your reply!Have you stained and normalised the patches?

jingweizhang-xyz commented 6 months ago

No.

Rokinluohhh commented 6 months ago

Hi,can you provide your code for creating patches?

jingweizhang-xyz commented 6 months ago

The patches come from a previous project from our co-authors around two years ago. I will contact them to see if they still have that code.

Rokinluohhh commented 6 months ago

ok,thans very much.And may I ask if you used cross validation?

jingweizhang-xyz commented 6 months ago

No, we use fold=0 as the validation set and repeat the experiment 3 times.

Rokinluohhh commented 6 months ago

ok Thanks for your patient reply!

Rokinluohhh commented 6 months ago

Hi,why you reshape there? 微信截图_20240118092851

jingweizhang-xyz commented 6 months ago

Because we use merged images. For 200 merged images of 224x 224, it shape is h=224*200, w=224, c=3. This reshaping operation convert it back to (200, 224, 224, 3).

Rokinluohhh commented 6 months ago

So,I should change this code if i use images that havent merged?

jingweizhang-xyz commented 6 months ago

No, you do not need to. This code is capable with images not being merged. The reshaping operation will result in (1, 224, 224, 3).

Rokinluohhh commented 6 months ago

Thanks a lot!