Erez Posner

[AK @_akhaliq]

Light Field Neural Rendering
abs: https://t.co/DlI2Zfq8Ej
project page: https://t.co/sbmqHRwZeH

[AK @_akhaliq]

Relightable Neural Video Portrait
pdf: https://t.co/YoMeLnjECJ
abs: https://t.co/rtrFGoDn6I

[AK @_akhaliq]

Open-World Entity Segmentation
pdf: https://t.co/3fREwaE73c
abs: https://t.co/PfRIjiNknI
github: https://t.co/1HbAlwnq8C

[Dave Zhenyu Chen @davech2y]

We released the code for Scan2Cap! Check it out here: https://t.co/DA7oKnqBtz

Big thanks to our team! @hexpheus @MattNiessner AngelXChang

[AK @_akhaliq]

Per-Pixel Classification is Not All You Need for Semantic Segmentation
pdf: https://t.co/lG6ZYV8XBp
github: https://t.co/bXqZ6pR3Fb
outperforms both current sota semantic (55.6 mIoU on ADE20K) and panoptic segmentation (52.7 PQ on COCO) models

[AK @_akhaliq]

TransformerFusion: Monocular RGB Scene Reconstruction using Transformers
pdf: https://t.co/N5Uj3gl6sK
abs: https://t.co/QV5rYbuGYM
project page: https://t.co/IoJKwt0pUg

[AK @_akhaliq]

Global Filter Networks for Image Classification
pdf: https://t.co/dIeGFqtllM
abs: https://t.co/48uTA872An
project page: https://t.co/LyAIupelxl
github: https://t.co/0BcTRgg4pJ

[Matthias Niessner @MattNiessner]

Social media is important to share research publicly and for junior researchers it's one of few ways to get yourself known.

So it's not a great move to take that away when conferences are mostly virtual - I believe the CVPR motion was a big mistake and should be re-considered.

[Lucas Beyer (bl16) @giffmana]

How @elonmusk becomes a Super Sayan, according to CLIP and StyleGAN

[Michael Black @Michael_J_Black]

TUCH estimates 3D human pose and shape with self-contact. We touch our bodies regularly, yet most regressors ignore this. With our #CVPR2021 oral and award-nominated paper, we address this with novel datasets and losses. Code is online: https://t.co/AhGerv2XgF

[Meshcapade @meshcapade]

#CVPR2021 is just around the corner! We’re so excited to be participating in the “SMPL made simple” tutorial, which will focus on the #SMPL model and how to use it.
For more info: https://t.co/kCXK9AqlbO

[Aljaž Božič @BozicAljaz]

We released the code of Neural Deformation Graphs paper.

Code: https://t.co/WtDZ7N6mwZ
Paper: https://t.co/7JtVFsgnPJ
Video: https://t.co/PbQyP7KbkT

Check us out this week at #CVPR2021 in Session 2!

@pablorpalafox @MZollhoefer @JustusThies @angelaqdai @MattNiessner

[Xiuming Zhang @xiuming_zhang]

NeRFactor is out! It's a physically-based model that factorizes appearance into shape and reflectance given just multi-view images under *one unknown* illumination (i.e., NeRF data). It supports free-viewpoint relighting (w/ shadows!) and material editing. https://t.co/X5p2itAp1T

[Daniyar Turmukhambetov @dantkz]

I'm happy to share our upcoming #CVPR #CVPR2021 paper "Single Image Depth Prediction with Wavelet Decomposition"
TLDR: monodepth but with wavelets for fewer decoder convolutions

arxiv: https://t.co/8JIsPX700a
code: https://t.co/e4p3FDLj1f

[Jon Barron @jon_barron]

Phew, mip-NeRF code release is done: https://t.co/5weTQppj9u. If you are playing with NeRF-like models, use this! Otherwise you'll have aliasing and sampling bugs that just won't go away. Plus, it's smaller (50%), faster (7%), and more accurate (17%-60%). https://t.co/JvI2WzYnaC

[Google AI @GoogleAI]

Reconstructing the 3D world from 2D video requires monocular depth estimation and video panoptic segmentation, but they are typically considered separately. Today we present ViP-DeepLab, which does both at once while achieving state-of-the-art performance→https://t.co/QSZsdiVdB1

[Ji Hou @j1h0u]

Excited to share our work Pri3D, learning 3D priors for 2D scene understanding tasks. Pre-training on 3D learns complementary features to ImageNet pre-trained model, leading to an 11.9% improvement on 20% data.
https://t.co/ep2thjGEhL
w/ @sainingxie ben @angelaqdai @MattNiessner

[Angela Dai @angelaqdai]

Neural Parametric for 3D Deformable Shapes -- a learned alternative to traditional parametric models that disentangles 4D dynamics into shape and pose, enabling fitting to depth sequences, interpolation, and retargeting!

@pablorpalafox @BozicAljaz @JustusThies @MattNiessner