Level 2 · CNNs & Image Classification

Classical filters are hand-designed; a convolutional neural network learns its own filters directly from real data. This level builds that understanding from the ground up: why convolution and pooling make sense for images, a real CNN trained from scratch and watched as it actually learns, real pretrained architectures (ResNet) used as backbones, real transfer learning and fine-tuning, and the evaluation and visualization tools that reveal what a trained model actually picked up on — not just its final accuracy number.

By the end of this level you'll have trained, evaluated, and fine-tuned real image classifiers, and built a complete, real image classification project end to end.