Usage instructions:
Training a profile:
- Create a folder for your training images. This folder should not contain anything except for the category folders and the images in them.
- Inside of your main images folder, create a folder for each category. Put all of your training images in its category’s folder. For example, if you are training a profile to distinguish between different flowers, it would look something like this:
(Don’t have a file called LICESNSE, just ignore that)
- Create a profile for this classifier in Taxon if you haven’t already
- Train your profile, and select flower_photos (or whatever yours is) as the image directory. Make sure that you don’t accidentally select a category folder or something when you do this. Before you press the train button, make sure the path shown on the screen is to the correct folder.
- Wait for the training to finish, and you’re ready to label!
How to use a trained profile:
- After you have trained a profile, press “use a profile”
- Select the profile you trained and select the image you want to label
- After a short time, it will tell you what category it thinks the image is and how confident it is.
Things you shouldn’t do:
- do not store any files in or manipulate the program’s directory in any way
For developers:
Dependencies (for non-frozen python):
- tensorflow version 1.5.0
- appJar version 0.90.0
- I have only found this to work on python 3.5.2 64 bit with tensorflow 1.5.0 installed