CIFAR-10 Image Classification — PyTorch Benchmark

An end-to-end image-classification project comparing a custom CNN, MobileNetV2 and ResNet-18 under controlled training and transfer-learning conditions. ResNet-18 achieved 87.48% test accuracy using transfer learning, outperforming the custom CNN baseline. The project includes Grad-CAM visual interpretability, INT8 quantisation, command-line inference tools and a live Gradio demo deployed on Hugging Face Spaces.