Pretrained Swin Transformer Models for Slide-Level Leukemia Classification
Authors/Creators
Description
This repository contains pretrained Swin Transformer model weights used for slide-level leukaemia classification from whole slide bone marrow aspirate images.
The models were developed as part of a two-stage pipeline:
(1) ROI-level classification to identify diagnostically relevant regions, and
(2) slide-level subtype classification using high-resolution patches extracted from ROI-positive areas.
The models are intended for inference and demonstration purposes and are used in conjunction with the accompanying application repository.
Due to file size limitations, the model weights are hosted separately from the main codebase. The corresponding implementation, preprocessing pipeline, and application can be found here.
Usage:
Download the model files and place them in the "models/" directory of the application repository before running the pipeline.
Note:
These models are provided for research and educational purposes only and are not intended for clinical use.
Files
Files
(543.2 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:da259c0554d56adab8aaf9f56fd10746
|
347.7 MB | Download |
|
md5:3c32f4f6d6dab0960c36aa701f3cc64b
|
195.5 MB | Download |
Additional details
Software
- Repository URL
- https://github.com/IngunnOfs/Swin-Transformers-for-Slide-Level-Leukaemia-Classification-with-Attention-Maps---Application
- Programming language
- Python