Walkthrough

This manual will explain the process of how to use Autotrace to trace images, train a network, and generate traces.

Tracing Manually

• Open Autotrace and follow the manual instructions.
  • If you do not already have traced data, it is recommended that you use this manual method to generate a training set upon which to construct the deep belief network as explained below.

Tracing Automatically

• Before anything, your images must be in .jpg format, not .png. If you have .png images, convert them using the included png_to_jpg.py script, by typing python png_to_jpg.py /path/to/png/images.

• First, use SelectROI to create a ROIconfig.txt file. Remember where you put it, we may need to move it later.

• Next, use ImageDiversity to separate a training and test set.

• Then, use ConfigDir to arrange the training data into the proper directory structure.

  • At this point, ROIconfig.txt should be in the /train/Subject1/ directory. If it is not, move it there.

• Now for the actual training. Follow the instructions in Train Network.

  • Ensure you point TrainNetwork at train/Subject1/ and train/traces using shift-click.
  • This may take a while. Be patient.
  • Now you have a Deep-Belief Network tailored to your data set that will be used by Autotrace.

• Now you are ready to use Autotrace. Open the manual and scroll down to the automatic tracing instructions.