Study notes that became the 3D ML atlas

What it does This repository is a curated reading list and dataset index for 3D machine learning, assembled from the maintainer’s own study notes. It catalogs courses, benchmark datasets like ModelNet and ShapeNet, and research papers across tasks including pose estimation, reconstruction, and scene synthesis. Each paper is tagged by its 3D representation—voxels, point clouds, meshes, primitives, or multi-view images—so you can browse by geometry type rather than hunting through scattered PDFs.

The interesting bit Instead of a flat bibliography, the repo uses emoji icons as a visual shorthand for data representations and maps tasks to specific 3D modalities. It also points readers to Connected Papers to visualize citation graphs, effectively turning a static markdown file into a lightweight literature map.

Key highlights

• Covers the full research arc: single-object classification, 3D pose estimation, semantic segmentation, geometry synthesis, texture analysis, and scene reconstruction
• Datasets range from legacy benchmarks (Princeton Shape Benchmark) to modern scanned and CAD-derived collections like ScanObjectNN and the Fusion 360 Gallery
• Papers are tagged by representation type: volumetric, point cloud, polygonal mesh, primitive-based, or multi-view images
• Includes active course links from Stanford, MIT, Princeton, and UCSD
• Maintains a Slack workspace for community questions and collaboration

Caveats

• The entire index lives in one massive README; finding a specific paper requires scrolling through manually curated sections with no built-in search or filter
• Many teaser images are hotlinked from external domains, so the visual experience depends on third-party hosting staying alive
• Contribution guidelines are minimal; the README invites pull requests but does not document inclusion criteria

Verdict If you are entering 3D machine learning and need a lay of the land, this is a solid map. If you need runnable code or a unified API, this is just a well-organized index—look elsewhere.