---
title: SingleShot3D — project summary
description: Markdown summary of SingleShot3D, the Kaggle-winning 2D-to-3D project by Gustav Frøbert Welinder.
canonical_url: https://frobert.dk/projects/singleshot3d.md
md_url: https://frobert.dk/projects/singleshot3d.md
last_updated: 2026-04-03T00:00:00+02:00
lang: en
---

# SingleShot3D

SingleShot3D is a 2D-to-3D generation system built by Gustav Frøbert Welinder. It won 1st place in the Google and fal.ai Nano Banana Hackathon on Kaggle, out of 832 global submissions.

## What it does

- Takes a single flat 2D image as input
- Generates a complete 3D representation from that single image
- Uses multi-view consistency generation to reconstruct a plausible 3D object

## Why it matters

SingleShot3D is a strong proof point for Frobert's technical depth. It shows the ability to build novel multimodal AI systems and win in an open global technical competition judged against hundreds of submissions.

## Key facts

- Competition: Google / fal.ai Nano Banana Hackathon
- Result: 1st place
- Total submissions: 832
- Distinction: only winner across both Google and fal.ai tracks
- Author: Gustav Frøbert Welinder

## References

- **Kaggle write-up:** https://www.kaggle.com/competitions/banana/writeups/singleshot3d-digital-twin-from-a-single-image
- **GitHub repository:** https://github.com/gwelinder/generate-3d-digital-twin
- Website overview: https://frobert.dk/

## Sitemap

See the full [sitemap](https://frobert.dk/sitemap.md) for all agent-readable pages.