Please checkout our new paper SceneCode presented in @cvpr2019. A compact code learned by a conditional VAE can control semantic segmentations and we can achieve coherent multi-view semantic label fusion by optimising the latent code.

Thank you Dr. Amani ! Really happy to share my PhD research to public audience.

We are looking for an outstanding new Dyson Research Fellow (super post-doc) in my lab, the Dyson Robotics Lab at Imperial College London. Please apply if you want to work on the future of #SpatialAI or manipulation for next generation robotics. jobs.ac.uk/job/BWI264/dys…

All researchers should fight against this. Every week I try to persuade my students that top papers often have few quantitative results. With work that's new, important, and clearly qualitatively different (zero to one!), you don't need quantitative results. Demos not tables!

Happy to introduce Semantic-NeRF. Multi-view consistency and smoothness make NeRF-training a label fusion process, supervised by sparse or noisy labels only! Work with: Tristan Laidlow, Stefan Leutenegger,Andrew Davison Project page: shuaifengzhi.com/Semantic-NeRF/ Paper: arxiv.org/abs/2103.15875

ReCo is a new contrastive learning framework for semantic segmentation. Guided by both local and global context, ReCo boosts (semi-)supervised methods by a large margin with few dense or several partial labels, revealing hierarchical similarities of various semantic classes.

We add semantics outputs to NeRF models of 3D occupancy/colour. Joint representation allows very sparse or noisy in-place supervision to generate high quality dense prediction. Dyson Robotics Lab #ICCV2021 Oral Shuaifeng Zhi Tristan Laidlow Stefan Leutenegger. shuaifengzhi.com/Semantic-NeRF/