fuad00 / bashnn

Example of a neural network writed on bash

BashNN

Example of a neural network writed on bash Its main purpose is to distinguish two different image types, like faces - no faces, etc.

Better source image resolution - better but slower learning.

To change the source model resolution, change number of steps in blankmodel() function to match the resolution multiplication, so 4096 for 64x64 as example.

Add the blankmodel function to main() to create the initial model, then remove to continue training.

Currently it will take .jpg images as negative examples and .png for positive to automate the training process, but you can uncomment and comment the check() function to match your needs.

Use ./evaluator.sh image.png to test with trained model