Repolode
Submit
humphd

humphd/have-fun-with-machine-learning

A hands-on, beginner-oriented tutorial for training and deploying image classification neural networks with Caffe and NVIDIA DIGITS, no prior AI background needed.

View on GitHub
Stars
5,117
Forks
530
Open issues
8
Last commit
2 months ago

What is it?

This is a practical guide to machine learning for programmers who have no AI background. It walks through building a classifier that tells dolphins from seahorses, using Caffe as the deep learning framework and NVIDIA's DIGITS web tool to manage datasets and training runs.

The guide covers the full workflow: preparing a categorized image dataset, training a convolutional network from random weights, then improving accuracy by fine-tuning pretrained AlexNet and GoogLeNet models. It shows how to change the final classification layers, interpret accuracy and loss, and export a model for deployment.

Final steps demonstrate running the trained model from the Caffe C++ example and a Python script that classifies new images. Setup instructions cover native installation and a Docker option. The material focuses on using existing tools rather than the underlying theory, and it points readers to external resources for that.

Key features

  • 01Walks through training a convolutional neural network on a small custom image dataset.
  • 02Shows fine-tuning pretrained AlexNet and GoogLeNet models for new classes.
  • 03Uses NVIDIA DIGITS for dataset creation, training charts and model testing.
  • 04Includes a Python script for classifying new images with a trained Caffe model.
  • 05Provides both native installation and a Docker setup path.

Use cases

  • Developers new to machine learning who want a hands-on image classification project.
  • Engineers learning transfer learning by adapting a pretrained network to their own images.
  • Anyone who wants to try Caffe or DIGITS without first studying neural network theory.

Quick start

  1. 1

    Install Docker if it is not already installed

  2. 2

    Pull and run the Caffe + DIGITS container, replacing the repository path with your clone location

    docker run --name digits -d -p 8080:5000 -v /path/to/this/repository:/data/repo kaixhin/digits
  3. 3

    Open http://localhost:8080 in your web browser

Things to know

  • Native installation of Caffe is difficult and can take days, especially on macOS.
  • The guide is built on Caffe and DIGITS, older tools that were written around 2016 to 2017.
  • The README does not state a clear open-source license (the license is listed as NOASSERTION).

In collections

Computer vision libraries & tools

Topics

#machine-learning#image-classification#neural-network#caffe#tutorial#transfer-learning#deep-learning

FAQ

What is have-fun-with-machine-learning?
A hands-on, beginner-oriented tutorial for training and deploying image classification neural networks with Caffe and NVIDIA DIGITS, no prior AI background needed. This is a practical guide to machine learning for programmers who have no AI background. It walks through building a classifier that tells dolphins from seahorses, using Caffe as the deep learning framework and NVIDIA's DIGITS web tool to manage datasets and training runs.
Is have-fun-with-machine-learning free to use? What license does it have?
have-fun-with-machine-learning is open source under the Other license. Check the license terms before using it in commercial products.
Is have-fun-with-machine-learning still maintained?
Updated within 6 months. The last commit to have-fun-with-machine-learning was on Sep 4, 2026.
What can I use have-fun-with-machine-learning for?
Developers new to machine learning who want a hands-on image classification project. Engineers learning transfer learning by adapting a pretrained network to their own images. Anyone who wants to try Caffe or DIGITS without first studying neural network theory.
How do I install have-fun-with-machine-learning?
Install Docker if it is not already installed Pull and run the Caffe + DIGITS container, replacing the repository path with your clone location: docker run --name digits -d -p 8080:5000 -v /path/to/this/repository:/data/repo kaixhin/digits Open http://localhost:8080 in your web browser
Where can I get have-fun-with-machine-learning?
The source code and installation instructions are on GitHub: https://github.com/humphd/have-fun-with-machine-learning

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects