SuperFit App: Asset Explorer

A companion web application for SuperFit — browse primitive assemblies, inspect fitting statistics, interactively edit primitives, and generate spatially-conditioned 3D meshes with SpaceControl.

Assembly Visualizer

Primitive-guided Generation

Features

• Browse & inspect — navigate asset folders, view per-iteration ResFit statistics, and scrub through the fitting timeline.
• Real-time shaders — view SuperFrusta assemblies as sphere-traced GLSL shaders (powered by SySL).
• Interactive editing — select individual primitives and modify their parameters live in the browser.
• Spatially-conditioned mesh generation — combine the edited primitive assembly with a text prompt and generate a detailed textured mesh via SpaceControl (Fedele et al.).

Install

Prerequisites

Install SuperFit and SpaceControl following their respective instructions.

Backend (Python)

pip install flask flask-cors

These are the only extra packages beyond what SuperFit needs.

Frontend (Node / Yarn)

cd asset_explorer_frontend
yarn install

Run

1. Start the backend

cd asset_explorer_backend
python app.py --location /path/to/asset/folders [--port 5000]

--location must point to a directory containing one or more asset subfolders, each with a primitive_assembly.pkl produced by SuperFit.

For SpaceControl mesh generation, set the SPACECONTROL_PATH environment variable:

export SPACECONTROL_PATH=/path/to/spacecontrol

2. Start the frontend (development)

cd asset_explorer_frontend
yarn dev

Open http://localhost:5173. API requests are proxied to the backend (default port 5000).

3. Production build

cd asset_explorer_frontend
yarn build

Then serve asset_explorer_frontend/dist/ and point it at the backend, or configure CORS accordingly.

BibTeX

If you use this application, please cite SuperFit and SpaceControl:

@misc{ganeshan2026superfit,
  title         = {Residual Primitive Fitting of 3D Shapes with SuperFrusta},
  author        = {Aditya Ganeshan and Matheus Gadelha and Thibault Groueix and Zhiqin Chen and Siddhartha Chaudhuri and Vladimir G. Kim and Wang Yifan and Daniel Ritchie},
  year          = {2026},
  booktitle     = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month         = {June},
}

@article{fedele2025spacecontrol,
  title         = {SpaceControl: Introducing Test-Time Spatial Control to 3D Generative Modeling},
  author        = {Fedele, Elisabetta and Engelmann, Francis and Huang, Ian and Litany, Or and Pollefeys, Marc and Guibas, Leonidas},
  journal       = {arXiv preprint arXiv:2512.05343},
  year          = {2025},
}

License

MIT. See LICENSE.