Snap’s Research team is kicking off the week in New Orleans at the 2022 Computer Vision and Pattern Recognition Conference. This year at CVPR, our team will share seven new academic papers, alongside the world’s leading researchers, that show breakthroughs across image, video, object synthesis and object manipulation methods.
We’ve worked closely with interns and external academic institutions on this work to make significant gains in video synthesis technology. These developments can ultimately inform what we bring to our community of Snapchatters around the world.
The work presented in our papers is based on the following developments: Our team has built implicit video representations, resulting in state-of-the-art video synthesis on a variety of tasks, while maintaining modest computational requirements. We then introduce two new problems in the domain: multimodal video synthesis and playable environments.
For example, the CLIP-NeRF paper was a collaborative research effort to study the manipulation of Neural Radiance fields. Neural Radiance fields make it possible to render objects using neural networks, without needing sophisticated graphics pipelines. Findings from this work can help inform improvements to the ways digital assets are created for use in augmented reality experiences. And, this PartGlot paper explores how machines can better understand shapes and objects around us using language models.
We’re excited about the potential of this work to unlock the creativity of our community and creators across our products and platforms in the future.
GOING TO CVPR?
Our team will be on site so come say hello! If you’d like to learn more about our papers, team, and products, stop by booth #1322 during the Expo (21 June - 23 June or email conferences@snap.com
2022 CVPR PAPERS
Written by and in collaboration with Snap Research
Playable Environments: Video Manipulation in Space and Time
Willi Menapace, Stéphane Lathuilière, Aliaksandr Siarohin, Christian Theobalt, Sergey Tulyakov, Vladislav Golyanik, Elisa Ricci Poster Session: Tuesday 21 June 2022 2:30PM – 5:00PM
Paper ID: 2345 | Poster ID: 99b
Show Me What and Tell Me How: Video Synthesis via Multimodal Conditioning Ligong Han, Jian Ren, Hsin-Ying Lee, Francesco Barbieri, Kyle Olszewski, Shervin Minaee, Dimitris Metaxas, Sergey Tulyakov
Poster Session: Tuesday 21 June 21 2022 2:30PM – 5:00PM
Paper ID: 3594 | Poster ID: 102b
CLIP-NeRF: Text-and-Image Driven Manipulation of Neural Radiance Fields
Can Wang, Menglei Chai, Mingming He, Dongdong Chen, Jing Liao Poster Session: Tuesday 21 June 2022 | 2:30PM – 5:00PM
Paper ID: 6311 | Poster ID: 123b
StyleGAN-V: A Continuous Video Generator with the Price, Image Quality and Perks of StyleGAN2
Ivan Skorokhodov, Sergey Tulyakov, Mohamed Elhoseiny
Poster Session: Tuesday 21 June 2022 | 2:30PM – 5:00PM
Paper ID: 5802 | Poster ID: 103b
Diverse Image Outpainting via GAN Inversion
Yen-Chi Cheng, Chieh Hubert Lin, Hsin-Ying Lee, Jian Ren, Sergey Tulyakov, Ming-Hsuan Yang
Poster Session: Thursday 23 June 2022 | 10:00AM-12:30 PM
Paper ID: 5449 | Poster ID: 79a
PartGlot: Learning Shape Part Segmentation from Language Reference Games
Ian Huang,Juil Koo, Panos Achlioptas, Leonidas Guibas, Minhyuk Sung
Poster Session: Friday 24 June 2022 8:30 AM - 10:18 AM
Paper ID: 3830 | Poster ID: 49a
Are Multimodal Transformers Robust to Missing Modality?
Mengmeng Ma, Jian Ren, Long Zhao, Davide Testuggine, Xi Peng
Poster Session: Friday 24 June 2022 | 10:00AM - 12:30 PM
Paper ID: 7761 | Poster ID: 212a