CSE 590V: Vision Seminar

Fall 2021

Course description

CSE 590V is a seminar/reading group focused on recent work in computer vision and graphics. We will cover papers from recent and upcoming conferences related to computer vision (CVPR, ICCV, ECCV, SIGGRAPH, NeurIPS, ICLR, ICML). We will also have guest lectures talking about their researches. The seminar is open to everyone. We especially encourage first year graduate students who may be considering research in computer vision or related areas to participate.


Logistics

Time: Mondays @ 11:00am - 12:00pm

Zoom link: Here

Organizers: Luyang Zhu (lyzhu @ cs washington edu) and Johanna Karras (jskarras @ cs washington edu)


Schedule

Date Presenters Papers Slides
October 4 Luyang/Yifan Fast NeRFs Slides
October 11 Chung-Yi 3D Generative Models
  • HoloGAN: Unsupervised learning of 3D representations from natural images
  • GRAF: Generative Radiance Fields for 3D-Aware Image Synthesis
  • pi-GAN: Periodic Implicit Generative Adversarial Networks for 3D-Aware Image Synthesis
  • GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields
Slides, Recording
October 18 Jingwei Image-based HMR
  • [ECCV 2016] Keep it SMPL: Automatic Estimation of 3D Human Pose and Shape from a Single Image
  • [CVPR 2018] HMR: End-to-end Recovery of Human Shape and Pose
  • [ICCV 2019] SPIN: SMPL oPtimization IN the loop
  • [ICCV 2021] Probabilistic Modeling for Human Mesh Recovery
  • [ICCV 2021] PARE: Part Attention Regressor for 3D Human Body Estimation
Video-based HMR
  • [CVPR 2019] Temporal HMR: Learning 3D Human Dynamics from Video
  • [CVPR 2020] VIBE: Video Inference for Human Body Pose and Shape Estimation
  • [ICCV 2021] HuMoR: 3D Human Motion Model for Robust Pose Estimation
Slides, Recording
October 25 Guest Speaker: James Tompkin
November 1 Meng-Li, Xiaojuan Recent Techniques on Holistic Video Editing:
  • Layered neural rendering for retiming people in video
  • Omnimatte: Associating Objects and Their Effects in Video
  • Layered neural atlases for consistent video editing
Slides, Recording
November 8 Johanna Generative Vision Transformers;
  • Zero-Shot Text-to-Image Generation
  • Taming Transformers for High-Resolution Synthesis
  • TransGAN: Two Pure Transformers Can Make One Strong GAN, and That Can Scale Up
  • Generative Adversarial Transformers
  • ViTGAN : Training GANs with Vision Transformers
Slides, Recording
November 15 CVPR DDL No seminar
November 22 CVPR Supp DDL No seminar
November 29 Benlin Editing Conditional Radiance Fields Slides, Recording
December 6 One-Shot Free-view Neural Talking-Head Synthesis for Video Conferencing