| --- |
| license: apache-2.0 |
| language: |
| - en |
|
|
| tags: |
| - cytopathology |
| - pathology |
| - medical-imaging |
| - foundation-model |
| - self-supervised-learning |
| - vision-transformer |
|
|
| library_name: pytorch |
| pipeline_tag: image-feature-extraction |
| --- |
| |
| # CROWN |
| A Universal Visual Foundation Model for Computational Cytopathology. |
|
|
|
|
| CROWN is a visual foundation model pretrained on over 10 million cytology images. |
| It provides transferable and annotation-free feature representations for a wide range of cytological image analysis tasks. |
|
|
| ### Highlights |
| - Pretrained with a DINOv2-style self-supervised framework on large-scale cytology data |
| - Strong transferability across classification, retrieval, segmentation, detection, and slide-level weakly supervised tasks |
| - Robust under significant domain shifts, serving as a single scalable backbone for cytology research |
|
|
|
|
| ## Quick usage |
| After downloading our model weights： |
|
|
| ``` |
| from models.vision_transformer import vit_large |
| import torch |
| |
| model = vit_large( |
| patch_size=16, |
| img_size=224, |
| init_values=1.0, |
| block_chunks=4, |
| ffn_layer="swiglufused", |
| ) |
| |
| state_dict = torch.load("CROWN.pth", map_location="cpu") |
| model.load_state_dict(state_dict, strict=True) |
| model.eval() |
| |
| ``` |
|
|
| Used for feature extraction: |
|
|
| ``` |
| feat = model.forward_features(img)["x_norm_clstoken"] |
| ``` |
|
|
| ## Citation |
| If you find our work useful in your research or if you use parts of this code please consider citing our paper: |
|
|
| ``` |
| |
| ``` |
|
|
