Generic-CNN-Project-with-Tensorflow

Classify images using CNN module from tensorflow.

Project setup

Install dependencies

1. install python & tensorflow

• beware of GPU driver's version & tensorflow's version.
• do make sure softwares are compatible. See https://www.tensorflow.org/install/source#gpu

2. and everything else

• Pillow
• matplotlib
• numpy

Running

$ python3 /path/to/classes.py

with tkinter, the GUI should look something like this:

Usage

Prepare your dataset

1. Each folder should contain multiple images.
2. Each folder represents 1 class.
3. Each folder's name will be used as class name.

Training

Simply select "Train" and click "Run".

Wait till the training is over.

Predicting

Select target directory containing images.

Click "Test" and then "Run".

Predictions are shown in the message box(top left).

Limitations

• Images are not scaled dynamically. Sampling height/width should equal to that of training.
• Images are not cleaned, anything from the background are used in training.
• "Training Checkpoints" not yet implemented, beware of machine shutting down during training.