Image Classifier's GUI which allows the user to change the network's training settings, see the program's progress in real time, and view any testing / training images.
Image Classifer is the first machine-learning program I've made, using Python's PyTorch library to train over the 10,000 images in the CIFAR-10 dataset, and creating a convolutional neural network that can identify sample images into one of 10 classes (all pictures are either of a plane, a car, a bird, a cat, a deer, a dog, a frog, a horse, a ship, or of a truck). The program loads the CIFAR-10 image dataset using PyTorch's Torchvision sub-library, using its small 32x32 pixel RGB images to test and train off.
Image Classifier's GUI which allows the user to change the network's training settings, see the program's progress in real time, and view any testing / training images.
The source code for the program can be copied from its GitHub repository HERE, and includes 3 main files: the program's main .py file called Image-Classifier.py,
the program's icon image called app.ico (this MUST be copied alongside the .py file for the program to run), and a pre-trained .pth neural network model called classifier_net.pth
(allowing users to instantly test a pre-trained model, if they so desire).
The program requires Python version 3.10 to run, with the PyTorch, Torchvision, Numpy (MUST be pre-version 2), and Matplotlib
libraries (alongside the default Python libraries of OS and Tkinter).
Due to the size of the PyTorch and Torchvision libraries, I decided not to turn the program into a .exe application for the time being, but may
still do this at a later stage.