-
Notifications
You must be signed in to change notification settings - Fork 0
CSU-BME/Pathology_SR
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
This is a super-resolution algorithm for pathology. The program runs on the mmcv_full==1.6.0 library. Here is the code for the manusript: Accurate diagnosis achieved via super-resolution whole slide images by pathologists and artificial intelligence with the doi: https://medrxiv.org/cgi/content/short/2024.07.05.24310022v1. (1) Dataset: A minimized MSR training and validation dataset, from which 10% of images were randomly sampled for each. Download link:https://pan.baidu.com/s/1Uu0P-NFX9W6NXlVGcvLA4g?pwd=g33c (2) Modify configs/restorers/Pathology_MSR/glean_in256out2048_pathology.py: Update paths in the data dictionary for the dataset: lq_folder='your path' gt_folder='your path' Set other hyperparameters inside as needed, including loss functions, number of iterations, etc. (3) configs/restorers/Pathology_MSR/create_anno.py: Select a subset of images from the validation set for validation (to reduce validation time) and generate the create_anno.txt file. (4) tools/train.py: Train the MSR model. Requires a Tesla A100 80GB GPU, approximately 3-4 days depending on the number of iterations. (5) demo/restoration_demo.py: Generate a 40X image from a 5X image for demonstration. (6) tools/generate_WSI.py: Generate 40X whole slide images from 5X image patches. (7) Pretrained models: pretrained/finetuned-style-gan.pth: The finetuned StyleGAN model for pathology images. pretrained/pretrained_MSR.pth: A pretrained MSR model. Download link: https://pan.baidu.com/s/1KXqGceto7GhH3f02RzjnXQ?pwd=qdax
About
This is a super-resolution algorithm for pathology. The program runs on the mmcv_full==1.6.0 library.
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published