Link Search Menu Expand Document

UnMICST-U (UNet)

Installation instructions

  1. Install python 3.7 and Anaconda
  2. Create a conda environment: conda create -n tf_gpu tensorflow-gpu=1.12 cudatoolkit=9.0
  3. Activate conda environment : conda activate tf_gpu
  4. Install the following packages: conda install scikit-image and conda install -c conda-forge opencv

Operation instructions

  1. Download Python scripts and ImageScience folder from code/UnMICST-U folder.
  2. Download models from dropbox models/UnMicst-U. https://www.dropbox.com/sh/3aqp83f5w1pxk0y/AABFgNRMJD2EvfSLFgCrXrBba?dl=0 Unzip and copy the subfolder to the same level as the script file(s).
  3. Download the training data from the training data dropbox folder. Unzip the folder to the same level as the script file(s).

Run model on new images by python <script name>.py train/test/deploy <append parameters below>

  1. function : whether to train a new model, test existing model on test data, or deploy the model on a completely new image.
  2. --imagePath : absolute path to an image you wish to deploy model on.
  3. --outputPath : specify where to save the output files after deploying model on new image.
  4. --channel : specify the channel(s) to be used. For DNA only models, only one channel should be specified. For DNA_NES models, use 2 channels ie. --channel 0 3
  5. --scalingFactor : an upsample or downsample factor if your pixel sizes are mismatched from the dataset.