qfnulxh / you-only-look-cytopathology-once

Codes available of a paper: An Efficient Cervical Whole Slide Image Analysis Framework Based on Multi-scale Semantic and Location Deep Features.

You-Only-Look-Cytopathology-Once

This is the code available of a preprint (under review) paper. Please cite it. arXiv:2106.15113

News:

• A new finding that the proposed InCNet of YOLCO has a very similiar structure of Hopfield network (1982), see following detailed figure of Hopfield network from Wikipedia.

• The speed is 55.9s for the inference of a WSI has 4,089,446,400 RGB pixels (63,949 x 63,949 x 3) in our cpp release version !

OS: Ubuntu 21.04
GPU: NVIDIA GeForce 1080Ti (11177 MB)
CPU: Intel(R) Core(TM) i7-7820X CPU @ 3.60GHz (16 cores)
RAM: 94 GB

• Example of feature maps can be found here: Colab

Project Dir Tree:

• Patch level classification cls-patch-level/
• Patch level detection det-patch-level/
• Qualitative results qualitative-results/
• WSI level classification cls-WSI-level/
• WSI level feature extraction det-WSI-level/
• Models available - See det-WSI-level/README.md
• C++ release version cpp/
• Data available

Graphic Abstract