Visual Atlas

Explore public domain image archives as navigable visual landscapes instead of searchable databases.

Visual Atlas is an experimental infinite-canvas interface that transforms large public image archives into navigable visual landscapes.

Instead of searching through lists, folders, or grids, users can freely pan across a continuously generated visual space where related images naturally cluster together. The result is a more intuitive, exploratory way of discovering connections, patterns, and forgotten collections within massive image archives.

It can also be used for your own personal collections — images, screenshots, and visual fragments that otherwise slumber forgotten on your hard drive — turning them into a living, explorable memory space.

The Problem

Large visual collections are searchable, but they are not explorable.Many public image collections contain millions of valuable images, but traditional search interfaces often limit discovery to keyword-based retrieval and static browsing.

Visual Atlas explores a different question:

What if archives could be explored spatially, like places rather than databases?

By organizing images through semantic similarity and presenting them on an infinite canvas, users can navigate archives through visual relationships and serendipitous exploration.

Links

Live Project: https://beneb85.github.io/visual-atlas/ Github: https://github.com/beneb85/visual-atlas Figma Community File: https://www.figma.com/community/file/1647949199129256120/visualatlas-ui-prototype Figma Make (only Canvas, without API Integrations) https://closed-foam-24445324.figma.site/

Process

The project combines AI-assisted development, semantic clustering, and rapid interface prototyping with Figma, Figma MCP, Figma Make and Claude

Development Workflow

I used Figma to create the initial design then used Figma MCP to design the spatial memory system in Claude and Figma Make to rapidly generate and iterate the interactive infinite canvas prototype and interface, and refined it into a working memory exploration system.

Inspiration

Original canvas interaction concept with Spotify API by Chakib Mazouni

The project was also inspired by Eduardo Lunardi's tutorial on building seamless infinite canvas experiences:

Tech Stack: Vanilla single-file JS · Three.js · MediaPipe Tasks-Vision · Web Audio · IndexedDB · weserv proxy

Music Into The Deep by Pamela Yuen (Free Music Archive) Licensed under a  Creative Commons CC BY

Sources

The visual material displayed in Visual Atlas originates from public archive collections and source institutions.

Art Institute of Chicago, The Met, NASA, Internet Archive, Smithsonian institution, Biodiversity Heritage Libary

Users are encouraged to consult the original source institutions for authoritative metadata, provenance information, and licensing details.

#ConfigMakeathon @figma