dair-ai/ml-visuals

ML Visuals is a free, community-built collection of diagrams and figures for explaining machine learning concepts. It is hosted as a Google Slides presentation and maintained through this GitHub repository. The goal is to give researchers, students, and writers ready-made visuals they can copy, customize, and use in blog posts, academic papers, and presentations, without having to draw neural network diagrams from scratch. The collection covers common ML concepts such as attention mechanisms, transformer architectures, encoder-decoder models, recurrent networks (RNNs, LSTMs, GRUs), and convolutional neural networks (CNNs), with more being added over time by community contributors. The README notes the slides already contain over 100 figures. All figures are free to use, the only ask is to credit the original designer, whose name is stored in the slide notes. Using it is straightforward: open the Google Slides link, browse the figures, and download whichever ones you need via File → Download in whatever format you want (PNG, PDF, etc.). If you want to customize a figure, you can request edit access to the shared document or make your own copy. Contributing is also encouraged, you add a new slide with your figure and include author information in the notes so others can credit you. You would reach for ML Visuals when you are writing a technical blog post, preparing a conference talk, or working on a paper and need a clean, clear visual of a standard ML concept but do not want to build the diagram from scratch in a drawing tool.