datasets: - wltjr1007/Camelyon17-WILDS
ViT-MIL for CAMELYON17-WILDS This model is a Vision Transformer (ViT) based Multiple Instance Learning (MIL) framework designed for detecting breast cancer metastasis in 96x96 pathches provided by the dataset titled as "wltjr1007/Camelyon17-WILDS" at Hugging Face dataset repository.
🔥 Reproducibility Training: https://github.com/kimdesok/Domain_Shift_on_Camelyon17/blob/main/ViT_MiL_on_Came17_wilds.ipynb
🏆 Institutional Achievement Developed as a pilot experimental model for the grant application such as National HPC Supporting Program by AICA, Gwangju, s.Korea or National NPU Support Program by NIPA, Jincheon, s. Korea This study model represents our commitment to reducing the manual workload of pathologists through high-performance AI.
📊 Model Details Architecture: ViT-Backbone with Multi-head Attention-based MIL Aggregator Training Data: wltjr1007/Camelyon17-WILDS Framework: Pytorch Target: Lymph node metastasis detection of breast cancers
📁 Dataset & Data Availability The model was trained on a curated version of the CAMELYON17 dataset, wltjr1007/Camelyon17-WILDS available at Hugging Face dataset repository.
Dataset Components: a. Patches: Extracted at 10.0x magnifications (for morphological features). *Type: <class 'PIL.PngImagePlugin.PngImageFile'> *Color schema: RGBA *Shape : (96, 96)
b. Sets: *Train set prepared from WSIs of three institutes with Center ID: {0, 3, 4} contains 302,436 patches. *Valid set of four institutes with Center ID: {0, 1, 3, 4} contains 68,464 patches. *Test set of one institute with Center ID: {2} contains 85,054 patches *All sets are 1:1 class-balanced with labels of {0, 1} *Columns are named as {'image', 'label', 'center', 'image_id', 'patient', 'node', 'x_coord', 'y_coord', 'slide'}
📈 Version History Version Date Description Status v1.0 2025-04-23 Initial Release (Fine-tuned on CAMELYON17-WILDS) Current v2.0 (TBD) Planned Virchow 2.0 Integration on H100 R&D Phase
⚠️ License & Commercial Use This model is licensed under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0).
Academics: Free to use for research and publications. Industry/Commercial: Use for-profit requires a separate commercial license. Inquiries: Please contact [dskim@btrust.co.kr] for licensing and collaboration.