This is the implementation code of 'A 3D Instance Segmentation Network for Synapse Reconstruction from Serial Electron Microscopy Images'.
The code is developed and tested under the following configurations.
• Hardware: more than 1 Nvidia GPUs with at least 24G GPU memory
• Software: Ubuntu 18.04, Keras=2.2.4, python=3.5
Adjust the input data path (include train data and test data, and point the save path for solving result) and run the main function. The result will be saved in specified path. If you want to example other datasets, please adjust the input data as the same as the above Test Data.
SNEMI data can be downloaded at: https://lichtman.rc.fas.harvard.edu/vast/AC3AC4Package.zip
This project is built upon numerous previous projects. Especially, we'd like to thank the contributors of the following github repositories:
Mask R-CNN: https://github.com/matterport/Mask_RCNN
pytorch_connectomics: https://github.com/zudi-lin/pytorch_connectomics
If you meet any problem, please contact the author directly. Email: [email protected]