!DOCTYPE html> SceneNet - A Library of Labelled Synthetic 3D Models

Repository of Labelled Synthetic Indoor Scenes. Make sure you have WebGL enabled in your browser to view the 3D models. We are also looking into generating unlimited data in the form of 3D scenes with our simulated annealing algorithm. We have also automated texturing of these scenes using archivetextures and opensurfaces.

We are increasingly seeing the use of these scenes beyond standard computer vision problems e.g. semantic segmentation, optic flow, 3D scene reconstruction etc. to now physical scene understanding and Deep Reinforcement Learning with agents interacting with their 3D environments.

Work in Progress https://github.com/ankurhanda/SceneNetv1.0

Living-room

Kitchen

Bedroom

Office

Bathroom



Big Scenes










Publications

SceneNet: Understanding Real World Indoor Scenes With Synthetic Data, 
A. Handa, V. Patraucean, V. Badrinarayanan, S. Stent and R. Cipolla
[up-to-date arXiv version of CVPR 2016 and ICRA 2016]
Understanding Real World Indoor Scenes with Synthetic Data,
A. Handa, V. Patraucean, V. Badrinarayanan, S. Stent and R. Cipolla, CVPR 2016
[pdf]
SceneNet: An Annotated Model Generator for Indoor Scene Understanding,
A. Handa, V. Patraucean, S. Stent and R. Cipolla, ICRA 2016
[pdf]

License

All the code and data are released under a creative commons license which is purely for research purposes only. Please view the summary here http://creativecommons.org/licenses/by-nc/4.0/.