KongNet_pretrained_weights

Model Card for KongNet

This repository contains pretrained weights for KongNet, a deep learning model for nuclei detection and classification.

Model Details

2025 MIDOG Challenge Model

• Developed by: Tissue Image Analytics (TIA) Centre, University of Warwick, United Kingdom.
• Model type: KongNet-Det
• License: CC BY-NC-SA 4.0
• Model Stats:
  • Params(M): 126M
  • Image Size: 512 x 512
  • Resolution: 0.25 mpp
• Cell Type:
  • Mitotic Figure
• Training Datasets:
  • MIDOG++
  • MITOS_WSI_CMC
  • MITOS_WSI_CCMCT
  • Mitosis Dataset for TCGA Diagnostic Slides
• Pretrained weights:
  • KongNet_Det_MIDOG_1.pth
  • KongNet_Det_MIDOG_3.pth
  • KongNet_Det_MIDOG_4.pth

CoNIC Model

• Developed by: Tissue Image Analytics (TIA) Centre, University of Warwick, United Kingdom.
• Model type: KongNet
• License: CC BY-NC-SA 4.0
• Model Stats:
  • Params(M): 176M
  • Image Size: 256 x 256
  • Resolution: 0.5 mpp
• Cell Type:
  • Epithelial Cell
  • Lymphocyte
  • Plasma Cell
  • Neutrophil
  • Eosinophil
  • Connective Tissue Cell
• Training Datasets:
  • CoNIC
• Pretrained weights:
  • KongNet_CoNIC_1.pth
  • KongNet_CoNIC_2.pth
  • KongNet_CoNIC_4.pth

MONKEY Challenge Model

• Developed by: Tissue Image Analytics (TIA) Centre, University of Warwick, United Kingdom.
• Model type: KongNet (wide)
• License: CC BY-NC-SA 4.0
• Model Stats:
  • Params(M): 174M
  • Image Size: 256 x 256
  • Resolution: 0.25 mpp
• Cell Type:
  • Overall mononuclear leukocyte
  • Lymphocyte
  • Monocyte
• Training Datasets:
  • MONKEY
• Pretrained weights:
  • KongNet_MONKEY_1.pth
  • KongNet_MONKEY_2.pth
  • KongNet_MONKEY_4.pth

PanNuke Model

• Developed by: Tissue Image Analytics (TIA) Centre, University of Warwick, United Kingdom.
• Model type: KongNet
• License: CC BY-NC-SA 4.0
• Model Stats:
  • Params(M): 176M
  • Image Size: 256 x 256
  • Resolution: 0.25 mpp
• Cell Type:
  • Overall
  • Neoplastic Cell
  • Non-Neoplatic Epithelial Cell
  • Inflammatory Cell
  • Dead Cell
  • Connective Cell
• Training Datasets:
  • PanNuke
• Pretrained weights:
  • KongNet_PanNuke_1.pth
  • KongNet_PanNuke_2.pth
  • KongNet_PanNuke_3.pth

PUMA Challenge Model (Track 1)

• Developed by: Tissue Image Analytics (TIA) Centre, University of Warwick, United Kingdom.
• Model type: KongNet
• License: CC BY-NC-SA 4.0
• Model Stats:
  • Params(M): 146M
  • Image Size: 256 x 256
  • Resolution: 0.25 mpp
• Cell Type:
  • Tumour Cell
  • Lymphocyte
  • Other Cell
• Training Datasets:
  • PUMA
• Pretrained weights:
  • KongNet_PUMA_T1_3.pth
  • KongNet_PUMA_T1_4.pth
  • KongNet_PUMA_T1_5.pth

PUMA Challenge Model (Track 2)

• Developed by: Tissue Image Analytics (TIA) Centre, University of Warwick, United Kingdom.
• Model type: KongNet
• License: CC BY-NC-SA 4.0
• Model Stats:
  • Params(M): 216M
  • Image Size: 256 x 256
  • Resolution: 0.25 mpp
• Cell Type:
  • Tumour Cell
  • Lymphocyte
  • Plasma Cell
  • Histiocyte
  • Melanophage
  • Neutrophil
  • Stroma Cell
  • Epithelial Cell
  • Endothelial Cell
  • Apoptotic Cell
• Training Datasets:
  • PUMA
• Pretrained weights:
  • KongNet_PUMA_T2_3.pth
  • KongNet_PUMA_T2_4.pth
  • KongNet_PUMA_T2_5.pth

Model Sources

<!-- Provide the basic links for the model. -->

• Inference Code (Github): KongNet_Inference_Main
• Paper: KongNet: A Multi-headed Deep Learning Model for Accurate Detection and Classification of Nuclei in Histopathology Images

Citation

If you use this model, please cite:

BibTeX:

@article{LV2026104233,
title = {KongNet: A multi-headed deep learning model for detection and classification of nuclei in histopathology images},
journal = {Medical Image Analysis},
volume = {114},
pages = {104233},
year = {2026},
issn = {1361-8415},
doi = {https://doi.org/10.1016/j.media.2026.104233},
url = {https://www.sciencedirect.com/science/article/pii/S1361841526003026},
author = {Jiaqi Lv and Esha Sadia Nasir and Kesi Xu and Mostafa Jahanifar and Brinder Singh Chohan and Behnaz Elhaminia and Shan E Ahmed Raza},
keywords = {Nuclei detection, Deep learning, PAS-stained images, H&E-stained images},
}