U-RISC-Data-Code

This repo holds data and code for U-RISC cell membrane segmentation.

Dataset

1. Download

There are three ways for users to download the dataset.

• Google Drive

https://drive.google.com/drive/folders/14u-TMK8kDJWMvZbr7KTQKwHpFqe9lrYU?usp=sharing

• MEGA

https://mega.nz/folder/HTIBkQzb#Dg2duJyMlghT9u5tM6EH7Q

• Baidu Cloud

https://pan.baidu.com/s/11WRPCJOu-R1h5ns3Fht80A code：zciq

2. Folder Directory

• U-RISC-DATA
  • imgs
    • Track1
      • train
      • val
      • test
    • Track2
    • Track2_original_size
  • label
  • Human_Annotation_Process
    • 1st
    • 2nd
    • final

Evaluation

The output of U-RISC dataset are binary segmentation images. And the evaluation criterion is F1-score.

Examples:

• Evaluate on Single Image:

   cd Evaluation
   python eval.py --evalon1pic True --pre_path ./pre.png --gt_path ./gt.png
  

• Evaluate on Forder Image:

   cd Evaluation
   python eval.py --evalon1pic False --pre_path ./pres --gt_path ./gts
  

Segmentation Networks

This part holds code for segmentation networks of U-RISC cell membrane segmentation.

Updating...

Citations

@article{shi2021u,
  title={U-RISC: An Ultra-high Resolution EM Dataset Challenging Existing Deep Learning Algorithms},
  author={Shi, Ruohua and Wang, Wenyao and Li, Zhixuan and He, Liuyuan and Sheng, Kaiwen and Ma, Lei and Du, Kai and Jiang, Tingting and Huang, Tiejun},
  journal={bioRxiv},
  year={2021},
}