have-fun-with-machine-learning

Why we included this project

This one is for programmers who have never trained a model and want to fix that without wading through theory first. The whole guide hangs off a single concrete goal: build a convolutional neural network that can tell dolphins from seahorses in photos it has never seen. The author deliberately skips the math-heavy first principles and works with Caffe and NVIDIA DIGITS instead, so you learn by doing, moving from building a dataset and training a network from scratch to fine-tuning pretrained models like AlexNet and GoogLeNet for better accuracy. Because it lives on GitHub and accepts pull requests, the author treats it as a living document that invites correction. If you want a low-friction way to get your hands dirty with image classification, this is a solid place to start.