Flash3D: Feed-Forward Generalisable 3D Scene Reconstruction from a Single Image
Authors:
Top Cited Papers
- 25 March 2025
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 670-681
Abstract
We propose Flash3D, a method for scene reconstruction and novel view synthesis from a single image which is both very generalisable and efficient. For generalisability, we start from a 'foundation' model for monocular depth estimation and extend it to a full 3D shape and appearance reconstructor. For efficiency, we base this extension on feed-forward Gaussian Splatting. Specifically, we predict a first layer of 3D Gaussians at the predicted depth, and then add additional layers of Gaussians that are offset in space, allowing the model to complete the reconstruction behind occlusions and truncations. Flash3D is very efficient, trainable on a single GPU in a day, and thus accessible to most researchers. It achieves state-of-the-art results when trained and tested on RealEstate10k. When transferred to unseen datasets like NYU it outperforms competitors by a large margin. More impressively, when transferred to KITTI, Flash3D achieves better PSNR than methods trained specifically on that dataset. In some instances, it even outperforms recent methods that use multiple views as input. Code, models, demo, and more results are available at https://www.robots.ox.ac.uk/~vgg/research/flash3d/.Keywords
All Related Versions
Funding Information
- Royal Academy of Engineering (RF1201819\18\163)
- EPSRC SYN3D (EP/Z001811/1)
This publication has 66 references indexed in Scilit:
- GRM: Large Gaussian Reconstruction Model for Efficient 3D Reconstruction and Generation2024
- LatentSplat: Autoencoding Variational Gaussians for Fast Generalizable 3D Reconstruction2024
- Factorizing Text-to-Video Generation by Explicit Image Conditioning2024
- MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-view Images2024
- LGM: Large Multi-view Gaussian Model for High-Resolution 3D Content Creation2024
- GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting2024
- Generalizable Patch-Based Neural Rendering2022
- SinNeRF: Training Neural Radiance Fields on Complex Scenes from a Single Image2022
- NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis2020
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015