Computer Vision for Cancer Cell Cycle Research
This Github contains the scripts developed to implement Convolutional U-Nets for automatic image analysis in cancer cell cycle research. The work was modeled on the DeLTA double U-Net analysis pipeline.
The python packages required to train the U-Nets can be installed from the YAML file in the Anaconda Packages folder.
To train a network set up the following folder environment with the data processing, model architecture and training scripts in the correct location:
--> scripts
-> data.py
-> model_seg.py
-> train_seg.py
--> data
-> img
-> truth
Data importing functions can currently deal with a mixture of 1024x1024 Barr Lab images as well as 696x520 BBBC database images. Images and their corresponding grouth truth masks must be saved with identical filenames in their respective folders. To train the U-Net run the training script from its location in the scripts folder.