This is a convolutional neural network implementation for garbage image classification
The dataset we are using is from the work by Gary Thung and Mindy Yang and can be found here
I have splitted the dataset into train and test set with the following directory structure:
.
+-- dataset-splitted
| +-- test-set
| | +-- cardboard
| | +-- glass
| | +-- ...
| +-- training-set
| | +-- cardboard
| | +-- glass
| | +-- ...
- tensorflow (or tensorflow-gpu if using GPU for training)
- keras
I suggest to run using the -i interpreter option to run commands on terminal
after execution such as fit(epochs)
or print_layers()
For this project I tried to take advantage of the transfer learning technique which turns out to be very promising. For the notes about the learning with different parameters see: densenet and VGG-19