https://github.com/cv516buaa/msrnet

https://github.com/cv516buaa/msrnet

Repository

https://github.com/cv516Buaa/MSRNet/blob/main/

# MSRNet

This repo is the implementation of "Look in Different Views: Multi-Scheme Regression Guided Cell Instance Segmentation". we refer to  [MMDetection](https://github.com/open-mmlab/mmdetection) to implement cell instance segmentation task. Many thanks to SenseTime and their excellent repos.



## Dataset
**2018 Data Science Bowl (DSB2018)** contains a total of 670 images, and the difficulty of this dataset mainly lies in the variety of image sizes, magnifications, imaging types and cell types. You can access to this dataset from [kaggle](https://www.kaggle.com/c/data-science-bowl-2018/data).

**CA2.5** consists of 524 fluorescence images of 512512 size, which contains some severely densely packed cell images with large differences in the brightness. You can access to this dataset from [CA2.5-Net Nuclei Segmentation Framework with a Microscopy Cell Benchmark Collection](https://link.springer.com/chapter/10.1007/978-3-030-87237-3_43).

**The Sartorius Cell Instance Segmentation (SCIS)** is from a Kaggles cell instance segmentation competition recently, which focus on neuronal cell instance segmentation. This dataset consists of a total of 606 images of 520704 size. You can access to this dataset from [kaggle](https://www.kaggle.com/competitions/sartorius-cell-instance-segmentation/data).

**CPM17** consists of 64 H&E stained histopathology images which contains a total of 7570 annotated nuclear boundaries.

**MoNuSeg** is a challenging multi-organ nuclei segmentation dataset, which contains 30 H&E stained histopathology images of 1000*1000 pixels and 21,623 individual annotated nuclei.

## MSRNet
### Install

1. requirements:
    
    python >= 3.7
        
    pytorch >= 1.5
        
    cuda >= 10.0
    
2. prerequisites: Please refer to  [MMDetection PREREQUISITES](https://mmdetection.readthedocs.io/en/latest/get_started.html); Please don't forget to install mmsegmentation with

     ```
     cd MSRNet
     
     pip install -e .
     
     chmod 777 ./tools/train.py
     
     chmod 777 ./tools/test.py
     ```

### Training

#### Task: Cell Instance Segmentation


  
     cd MSRNet
     
     python tools/train.py configs/msrnet/msrnet_1x_coco.py


### Testing

#### Task: Cell Instance Segmentation
  
     cd MSRNet
     
     python tools/test.py configs/msrnet/msrnet_1x_coco.py checkpoints/dsb2018_fin4_770_622.pth --eval bbox segm


## Description of MSRNet
- submitted to BHI2023

If you have any question, please discuss with me by sending email to lyushuchang@buaa.edu.cn.

# References
Many thanks to their excellent works
* [MMDetection](https://github.com/open-mmlab/mmdetection)

Owner

GitHub Events