The model uses PyTorch and CUDA compatible GPUs to train through 30 epochs, ultimatley acheiving ~90% accuracy. Example outputs: