A neural network library built without any AI libraries, only NumPy.
You have to set any images you want the model to train off of in the train/ folder.
Then, run src/ptl-neural-network/train.py. It will automatically take all the images and train the A.I. model. This could take a while!
It will store the output as a model.pkl file.
You can now run src/ptl-neural-network/run.py, with the only requirement being the model.pkl that you got from training.
It will ask for a .png file, and you can provide a relative or absolute path here. It will then find the closest image that matches your PNG.
Note that its output is the name of the image found from training data, so you might have to do some filtering to determine what the PNG is (bus) instead of a image name (bus3)
You can also do pip install ptl-neural-network to get the CLI.
The command syntax is ptl-neural train/run. The above guide also works with the CLI, except don't run train.py or run.py.
