spandan-madan/deeplearningproject

★ 4,778HTMLAudience · generalComplexity · 3/5Setup · moderate

Mindmap

mindmap
  root((deeplearningproject))
    What it does
      ML project tutorial
      Dataset collection
      Model training
    Tech Stack
      Python
      PyTorch
      Jupyter
      Docker
    Use Cases
      Learning ML end-to-end
      Building custom datasets
    Audience
      ML beginners
      Students

mindmap root((deeplearningproject)) What it does ML project tutorial Dataset collection Model training Tech Stack Python PyTorch Jupyter Docker Use Cases Learning ML end-to-end Building custom datasets Audience ML beginners Students

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Work through a full machine learning project end-to-end to understand what the real process looks like beyond simplified toy examples.

USE CASE 2

Build and label your own dataset from scratch instead of using a pre-packaged one, following the tutorial's guidance.

USE CASE 3

Learn how to apply deep learning with PyTorch to a custom dataset you have assembled yourself.

USE CASE 4

Run the tutorial in a Docker container to avoid dealing with Python dependency conflicts on your local machine.

Tech stack

PythonJupyterPyTorchDockercondaHTML

Getting it running

Difficulty · moderate Time to first run · 1h+

Requires conda for environment setup, some older library versions have known compatibility issues documented in the README with workarounds.

In plain English

This repository is an end-to-end machine learning tutorial originally designed as a class project for a graduate data science course at Harvard University in 2016. Unlike short tutorials that skip over the messy parts, this one walks through the entire process that a real machine learning project involves, from collecting and building a dataset from scratch to training a deep learning model on it. The tutorial deliberately avoids standard practice datasets like MNIST (handwritten digits) that are commonly used in beginner examples. Instead, it guides you through assembling your own dataset, which is what you would actually have to do when working on a real problem. From there it covers conventional machine learning approaches before moving into deep learning, a category of techniques that use layered neural networks to find patterns in data. The content is delivered as an interactive Jupyter notebook, which is a format that mixes explanations, code, and output in a single document you can open in a browser. A version using the PyTorch framework, a popular tool for deep learning research, was added in 2018. You can also read the tutorial as a static HTML page without setting up any software. Setting up the code requires Python and a package manager called conda. The repository includes a configuration file that installs all the required libraries in one command. There is also a Docker option for running the notebook in an isolated container if you prefer not to modify your local Python setup. The README notes some known compatibility issues between older versions of certain libraries and includes workarounds. This project is a learning resource rather than a reusable software library. Its audience is students and people new to machine learning who want a thorough walkthrough of what the full process looks like beyond the simplified examples found in most introductory content.

Copy-paste prompts

Prompt 1

I am following the spandan-madan/deeplearningproject tutorial. Walk me through how to collect and label my own image dataset for the machine learning section.

Prompt 2

Using the PyTorch notebook from this Harvard ML tutorial, help me adapt the training loop to work with my own dataset of images instead of the one in the tutorial.

Prompt 3

I am setting up the deeplearningproject tutorial with conda. Help me create the environment from the config file and troubleshoot any version conflicts.

Prompt 4

Explain the difference between the conventional machine learning approach and the deep learning approach shown in this Harvard tutorial, in plain terms.

Open on GitHub → Explain another repo

← spandan-madan on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.