Elliott / Shangzhe Wu
@elliottszwu
Followers
5,149
Following
790
Media
75
Statuses
515
Postdoc @StanfordSVL working on unsupervised 3D perception and inverse rendering, PhD from @Oxford_VGG . Public office hours:
Joined March 2015
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Leyen
• 281762 Tweets
梅雨明け
• 216960 Tweets
#BKPPยอม_MV
• 197379 Tweets
norawit x bv in emporium
• 115084 Tweets
INTRO
• 78744 Tweets
Papa Dali
• 54762 Tweets
Ali Koç
• 52834 Tweets
Kamari
• 51663 Tweets
パワプロ
• 48970 Tweets
MEW IN FRED SEIBU
• 48881 Tweets
岸田総理
• 36067 Tweets
VLOG 5 WITH BUILD
• 33358 Tweets
#素のまんま
• 33238 Tweets
डिब्रूगढ़ एक्सप्रेस
• 29433 Tweets
Mehmet Büyükekşi
• 26357 Tweets
रेल मंत्री
• 23634 Tweets
#TrainAccident
• 23164 Tweets
AMIA
• 21621 Tweets
国立公園
• 17187 Tweets
マンション爆発
• 16897 Tweets
高級ホテル誘致
• 14427 Tweets
What's your color
• 13521 Tweets
#THE夜会
• 13277 Tweets
高級リゾートホテル誘致
• 13143 Tweets
#ミュージックジェネレーション
• 11945 Tweets
西川口の爆発
• 11804 Tweets
Kimiko
• 10799 Tweets
Pinned Tweet
3D-Fauna will be presented at
#CVPR2024
on Thursday morning (10:30 - 12 poster
#11
)
Check out the 🤗 demo:
Great work led by
@zizhang_li
@DorLitvak
@StanfordSVL
& collab w/
@Oxford_VGG
🦌🦒🐫🦛🦣🐎🐃🐆🐕🐿️🐇🐖
Nature presents a captivating confluence of similarity and diversity.
Our new method 3D-Fauna learns a pan-category articulated 3D model of quadruped animals from Internet photos. At test time, it turns a single image into an animatable textured 3D mesh in a feed-forward pass.
2
35
213
0
9
69
🎠🎠 We're presenting MagicPony at CVPR!
It turns a photo of an animal into a fully articulated 3D model ready for animation. It is trained using just online images, without any 3D data, keypoints or template shapes.
Drop by our poster on Wed morning!
11
77
370
Let's turn photos of ancient "revolutionary" (rotationally symmetric) artefacts into 3D and rotate them, or even change the lighting!
Our model learns to de-render a single image of a vase into shape, albedo, material & lighting, from just a single-image collection.
#CVPR2021
4
39
257
I have passed my DPhil viva (defense) 🎓!
Four years flew by 💨 It is very sad to be leaving this wonderful place and so many lovely people
@Oxford_VGG
😢
Thank you the Best Advisors Andrea and
@chrirupp
who have shaped me into who I am today!
27
3
255
🕊️ DOVE: Learning Deformable 3D Objects by Watching Videos
Our method learns articulated 3D birds from only monocular videos, with automatically obtained masks and optical flow.
No keypoints, camera viewpoint or template shapes required!
Check the demo:
4
34
216
Nature presents a captivating confluence of similarity and diversity.
Our new method 3D-Fauna learns a pan-category articulated 3D model of quadruped animals from Internet photos. At test time, it turns a single image into an animatable textured 3D mesh in a feed-forward pass.
2
35
213
Super excited to announce that our workshop proposal on Unsup3D: Unsupervised 3D Learning in the Wild has been accepted to
#ICCV2021
!
We're putting together an incredible line-up of talks! Stay tuned!
@ICCV_2021
Congrats to the team
@chrirupp
@NataliaNeverova
Paolo & Andrea
0
26
160
Announcing
@ICCV_2021
workshop on "Unsupervised 3D Learning in the Wild" with an incredible line-up of speakers on this topic!
#ICCV2021
🚩 Website:
📅 Time: 7:00-18:00 EDT / 12:00-23:00 BST, 11 Oct 2021
Calendar: (mark it down!)
2
27
161
Code released for our
#CVPR2020
oral paper on unsupervised single image 3D reconstruction!
We also host a demo website:
Visualize your faces or cats in 3D!
w/
@chrirupp
, Andrea Vedaldi
@Oxford_VGG
2
45
150
🐎 Let the hooves pound!
Our new method Ponymation learns a generative model of 3D articulated animal motions from raw unlabeled Internet videos.
Page:
Paper:
Led by
@skq719
& Dor Litvak, w/
@zhang_yunzhi
Hongsheng Li
@jiajunwu_cs
3
21
122
If you're in the niche NeRF / 3D reconstruction community, consider participating in the CO3D Challenge!
Open until Oct 20:
hosted with the Neural Geometry and Rendering workshop
@eccvconf
0
17
91
Same as last year, I'll be sharing some hidden gems I found
#CVPR23
.
The first one is this really cool dataset of 3D broken objects 💔
It would be super cool if we could analyze the fracture patterns and infer fine-grained materials, or detect defects!
2
8
77
Latest trend in "self-supervised learning":
Abstract: we present a self-supervised method for ...
Somewhere on page 6: actually, we find it hard to train, so we pretrain on a SMALL set of labeled data
2
4
76
The Fundamental Matrix song by the creator Daniel Wedge and the mind-blowing
@CVPR
house band!
@yalesong
: "can you imagine Andrew Zisserman coming up to the stage and sing the song?"
@Oxford_VGG
Enjoy the full song:
2
9
73
Here are the video recordings of the workshop:
Announcing
@ICCV_2021
workshop on "Unsupervised 3D Learning in the Wild" with an incredible line-up of speakers on this topic!
#ICCV2021
🚩 Website:
📅 Time: 7:00-18:00 EDT / 12:00-23:00 BST, 11 Oct 2021
Calendar: (mark it down!)
2
27
161
0
12
72
Glad to share the TPAMI journal extension of our CVPR2020 paper:
We added many interesting experiments providing insights on how the method works in various scenarios, as well as evaluations on 3D face benchmarks.
0
6
70
Hello
@zhengqi_li
!
Amazing work and amazing slides by the team, and amazing talk by
@Jimantha
!
3
4
65
Let the hooves pound and the wings flap!
Finally found some time to clean up the code for both 🎠 MagicPony and 🕊️ Dove:
(sadly these birds haven't grown wings just yet..)
🎠🎠 We're presenting MagicPony at CVPR!
It turns a photo of an animal into a fully articulated 3D model ready for animation. It is trained using just online images, without any 3D data, keypoints or template shapes.
Drop by our poster on Wed morning!
11
77
370
0
10
64
Cool papers I found
#CVPR2021
and didn't see on twitter 👇
"HistoGAN: Controlling Colors of GAN-Generated and Real Images via Color Histograms" by Afifi et al.
3
4
56
We have released the code for our SoR (Solids of Revolution) De-rendering paper 🪄💡🏺
Let's turn photos of ancient "revolutionary" (rotationally symmetric) artefacts into 3D and rotate them, or even change the lighting!
Our model learns to de-render a single image of a vase into shape, albedo, material & lighting, from just a single-image collection.
#CVPR2021
4
39
257
0
9
54
The Neural Geometry and Rendering Workshop is starting in 35 mins in Salon A
#ECCV2022
!
1
8
52
Following in
@jbhuang0604
's footsteps, I've set up some public office hours:
I might not be of much help, but will always be happy to chat about research (or anything really) and learn from each other.
Keep conversations going and perhaps even further!
Open Office Hours
I set aside some time each week to meet with *anyone* (preferably students from underrepresented groups tho).
Happy to chat about any topics that may be helpful.
10-min for each slot, but you can reserve multiple slots if needed.
5
11
136
0
3
51
.
@ak92501
is real! Come to hall C!
@CSProfKGD
@elliottszwu
@CVPR
thanks, would also be great to meet, sent a dm, also I am at the coffee shop outside Hall C for the next 30 mins but will be leaving afterwards, will also be here tomorrow. I am wearing a green shirt
1
0
17
3
3
50
@Sagar_Vaze
It was you
When you keep getting the same reviewer in every resubmission
23
217
3K
0
0
44
This is happening TODAY (in ~7 hours)!
Join us via Zoom from
@ICCV_2021
platform! YouTube livestreaming will be put up shortly!
Announcing
@ICCV_2021
workshop on "Unsupervised 3D Learning in the Wild" with an incredible line-up of speakers on this topic!
#ICCV2021
🚩 Website:
📅 Time: 7:00-18:00 EDT / 12:00-23:00 BST, 11 Oct 2021
Calendar: (mark it down!)
2
27
161
0
8
38
Second talk by
@akanazawa
"Towards Dynamic 4D Capture -- Practices and Recommendations"
3
1
37
It was lovely to meet new vision friends around NYC!
Organizing the first NYC vision workshop was super fun! Shout out to other organizers
@elliottszwu
@Haian_Jin
and especially
@Jimantha
for the generous support!
3
8
65
0
0
37
#icvss2019
Jitendra Malik gave a great talk
@icvss
covering the history of CV, with a substantial focus on unsupervised 3D. Our recent work recovers 3D from single images without even keypoints/masks, using the idea of shape from shading. Poster
#72
on Tue
0
12
33
This is wild - the new NeRF song just dropped!!
14 years after the first release of the Fundamental Matrix Song:
Here's yet another
#NeRF
video... but this one is of the musical variety. Enjoy!
#WeAreTheNeRFs
1
23
103
0
3
32
Thanks AZ and
@vincesitzmann
for making it such an enjoyable viva!
And to my collaborators, labmates, friends, it's been an honor to have you!
Next step will be Stanford as a postdoc working with
@jiajunwu_cs
! Very sad departure but also a very exciting journey ahead!
0
1
28
"Self-Portraits"
Pareidolic faces are images in natural world on which we subconsciously cast a humanoid interpretation. These are moments we find ourselves in an unexpected encounter with another being. They are mirrors through which we see the images of ourselves as humans.
1
3
26
Had a great time visiting
@GRASPlab
and giving a talk on 3D Fauna and Flora!
Thanks
@LingjieLiu1
for hosting!
1
2
27
#ScholarPhi
is what I've been looking for years!
It peeps into symbol definitions, terms, and references inline in PDFs. Don't need to scroll all the way down to figure out the reference and forget where I was any more!
Demo looks amazing
Today at
#CHI2021
, I will be presenting a new project, "Augmenting Scientific Papers with Just-in-Time, Position-Sensitive Definitions of Terms and Symbols". You can catch it at 4pm PST today, 8am tomorrow, or at
Summary follows (1/8)
1
8
72
1
6
26
Really amazing talks in the first session! Thanks for the speakers! The talks have been recorded and will be released on the website later.
Announcing
@ICCV_2021
workshop on "Unsupervised 3D Learning in the Wild" with an incredible line-up of speakers on this topic!
#ICCV2021
🚩 Website:
📅 Time: 7:00-18:00 EDT / 12:00-23:00 BST, 11 Oct 2021
Calendar: (mark it down!)
2
27
161
2
3
25
Happy to be on the list of Top Reviewers
#NeurIPS
Reviewer awards have been one of the most mysterious ones. It's such a hard RL task with extremely sparse rewards. I'm very curious about the results of the review evaluation finally introduced this year.
0
0
24
So many great 3D papers in one single session (
#5
)
@ICCV_2021
and I could only attend so few of them!!
Very grateful some authors are still there even after 30 mins!
0
0
23
🌹 "People where you live ... grow five thousand roses in one garden ... And yet what they’re looking for could be found in a single rose." -- "The Little Prince" by Antoine de Saint-Exupery
Drop by our poster
#CVPR2023
on Tuesday morning!
What is a rose🌹visually? A rose is characterized by its intrinsics: the distribution of geometry, texture, material.
We learn a generative model that discovers object intrinsics from just a few instances in a single image - a task humans can do with ease.
#CVPR23
poster Tue AM
5
35
227
0
3
22
Super excited about this
#ECCV
workshop on Neural Geometry and Rendering and the CO3D Challenge!
Give it a try:
0
6
22
In a nutshell, we exploit the rotational symmetry of these vases to learn the 3D shape, and introduce a Self-supervised Albedo Discriminator to recover realistic albedo under specular reflections.
See the full video:
1
2
20
Check out the demo:
More results on the website:
Great work led by
@zizhang_li
&
@DorLitvak
with the amazing team
@RayLi234
@zhang_yunzhi
@JakabTomas
@chrirupp
Andrea Vedaldi
@jiajunwu_cs
PS:
@zizhang_li
is applying for PhD!!
0
4
21
The long envisioned life of a Doctor of Philosophy student. What happened to the past three years...
#PhDChatter
#postDeadlines
1
0
20
For CVPR folks at Vancouver, this jazz bar is really good!!!
Guilt & Co, 1 Alexander Street, Underground, Vancouver
0
1
20
The binary classification model for paper acceptance.
by
@greg_mori
at
#CVPR2020
Workshop on How to write a good review?
#makeReviewsGreatAgain
1
3
20
@akanazawa
This is so great! I remember having to memorize thounsands of words that I never encounter afterwards and to pay them ~30 bucks just to send the scores to each school electronically 🤷♂️
1
0
18
Huge thanks to my awesome mentors during my internship
@GoogleAI
:
@kiamada
@jiajunwu_cs
@Jimantha
Richard Tucker
@akanazawa
!!!
🚩
📄
@Oxford_VGG
0
4
18
Recent work with
@chrirupp
and Andrea Vedaldi on unsupervised learning of 3D objects. The model trained on real face images generalizes surprisingly well to paintings and abstract faces.
Video:
Learning weakly symmetric 3D objects from single-view images only, without any additional supervision, eg. multi-views, keypoints, or prior shape models etc, by
@elliottszwu
,
@chrirupp
and Andrea Vedaldi.
Paper:
Project:
0
16
72
0
4
16
📢 This is going live in 30 mins!
Join the Zoom meeting via
@ICCV_2021
platform.
YouTube live-streaming here:
Announcing
@ICCV_2021
workshop on "Unsupervised 3D Learning in the Wild" with an incredible line-up of speakers on this topic!
#ICCV2021
🚩 Website:
📅 Time: 7:00-18:00 EDT / 12:00-23:00 BST, 11 Oct 2021
Calendar: (mark it down!)
2
27
161
0
2
14
Surprised to find
#CVPR2021
cemera ready deadline to be Mar 26 (), which is 2 days after
#ICCV2021
sup. mat. deadline.
This sounds pretty stressful. Is it finalized?
@georgiagkioxari
0
0
14
New
#RSS2020
paper by
@TimYQTang
and coauthors on self-supervised cross-modality localisation.
We leverage cycle consistency for learning cross-modality SE(2) registration without GT pose or pose-aligned data, which allows vehicles to use satellite images for localisation.
[1/3] Check out our RSS2020 paper “Self-Supervised Localisation between Range Sensors and Overhead Imagery” by
@TimYQTang
, Daniele De Martini,
@elliottszwu
&
@paul_m_newman
.
video:
ArXiv:
#robotics
#rss2020
#AutonomousVehicles
1
7
9
0
2
14
This is such a great talk by
@Jimantha
! It's just so scientifically intriguing and inspiring that I couldn't stop looking into the mirror!
Highly recommend:
0
0
13
Thanks to the inflation, finally got it! 😆
0
0
13
1
2
13
We're deeply indebted to all members of
@Oxford_VGG
for insightful discussions and suggestions! A shout-out to a few: Sophia Koepke,
@gulvarol
@erika_lu_
@NagraniArsha
@oliviawiles1
@ftmguney
@irocp1
@danxuhk
@TengdaHan
@WeidiXie
@omgroth
@SamuelAlbanie
, Andrew Zisserman ...
1
0
12
First talk by
@BenMildenhall
"Modeling Light for View Synthesis"
0
1
12
🏆 Submit your results to the CO3D Challenge by 16:59PM PDT, Oct 22, 2022 and this could be yours!
1
3
11
This paper analyzing physical impacts from transient surface wave imaging is also super cool!
It essentially allows you to localize impact from tiny surface vibrations observed behind the direct line of sight.
Great work by
@Desmond_pku
@MarkSheinin
et al
1
1
11
✈️ I'll be around Bay Area between June 9-18 and would love to meet people. PM if you'd like to grab coffee ☕️!
2
0
10
If you're at
#NeurIPS2022
, feel free to say hi to
@skq719
who's presenting our recent work on controllable 3D face synthesis!
Excited to share our
#NeurIPS2022
work, cGOF, a NeRF-based conditional 3D face synthesis framework, which enables 3D controllability over the generated face images.
Poster:
📍Hall J
#240
⏰Tue 29 Nov 9 a.m. PST — 11 a.m. PST
Project:
2
3
11
0
0
10
@ursNwaldmann
@gengshanY
@annazama
Please consider contributing as a reviewer if you work on relevant fields of computer vision, biology, ecology, animal studies etc.:
Having a diverse set of expertise across various fields is absolutely crucial in this interdisciplinary topic.
0
1
10
As I'm finishing up the
#ICCV
reviews, I find this answer to the reproducibility question a bit awkward. I do not think that one line footnote of code promise makes me more believe its reproducibility.
Also, I miss the short rating justification. Now I add it to summary.
0
0
10
🏆 Congratulations to Taewan Ethan Kim on winning the the 1st place in the CO3D challenge on the "Many-view" track, and to
@zhizdev
on winning the 1st place in the "Few-view" track!
1
4
9
Second session with another 6 great talks!
Thanks to all the speakers for making it such a great event:
@talidekel
@adel_ahmadyan
Andreas (
@AutoVisionGroup
)
@akanazawa
@davnov134
Iasonas, Jitendra, Jingyi, Niloy
@Jimantha
@shalinidemello
@shubhtuls
@talidekel
!
1
1
8
@jbhuang0604
Aha they've evolved - group meetings. Interesting! How's that going?
I've had lots of fun with 1:1 conversations with such public hours. Curious to know if these group ones are even more fun.
1
0
8
@gulvarol
I guess referencing your paper itself might also work just fine, it should satisfy the request
0
0
8
Next talk is by Andreas Geiger (
@AutoVisionGroup
) "Constraining 3D Fields for Reconstruction and View Synthesis"
1
1
7
@wjscheirer
I'm really curious to see how Motion 4 will be implemented. I voted yes (for its spirit), but I think there are many challenges in terms of implementation, which the proposal did not address.
1
0
6
"Jigsaw Clustering for Unsupervised Visual Representation Learning"
by Pengguang Chen, Shu Liu, Jiaya Jia
1
0
7
@arkosiorek
I actually find it much easier to skim through the posters than watching the videos, especially when going though hundreds of papers and trying to decide in seconds whether to jump into the discussion or not.
1
0
7
Here are some comparisons with existing methods.
They use keypoints, camera viewpoint, template shapes, or (self-supervised) part segmentations for supervision.
0
0
6
I really like Shane Wighton's videos. They're like scientific reports, including motivation, methodology, results and ablations/backward analysis, but much more entertaining and also insightful.
It would be super cool to present papers in such a way 🙂
0
0
6
We will have an exciting line-up of talks across various disciplines to be announced soon (stay tuned)!
Co-organized with
@ursNwaldmann
@gengshanY
@annazama
and an amazing advising team!
1
0
6
0
0
6
We will have an amazing series of talks from
@adel_ahmadyan
Andreas (
@AutoVisionGroup
)
@akanazawa
@davnov134
Iasonas, Jitendra, Jingyi, Niloy
@Jimantha
@shalinidemello
@shubhtuls
@talidekel
!
Kudos to the great organizing team:
@chrirupp
@NataliaNeverova
Paolo & Andrea
0
0
5
Humans as Light Bulbs: reconstructing 3D humans from infrared radiation reflected off object surfaces
by
@ruoshi_liu
@cvondrick
Accidental Light Probes: recovering environment lighting from Coke cans
by
@Koven_Yu
and the team
1
0
5
Artwork by FrānkǎnLīsà ()
Visualizing pareidolic faces in 3D using our unsupervised machine learning model trained on human face images:
Preview version:
0
0
5
The unspoken truth
Have you ever wondered why papers from top universities/research labs often appear in the top few positions in the daily email and web announcements from arXiv?
Why is that the case? Why should I care?
21
159
840
1
0
5
A few more screenshots..
See you in 30 mins for the second session!
1
2
5
3D-Fauna leverages pre-trained unsupervised image features from DINO and automatically learns a semantically-adaptive bank of base 3D shapes for all quadrupeds in a canonical space, from scratch, without relying on any category labels.
1
0
5
joint work with Ray,
@JakabTomas
,
@chrirupp
and Andrea
@Oxford_VGG
Check out the video here (with great sound track!):
0
0
5
Also obviously the unspoken rule is that every CVPR paper must contain pictures of animals. If this is news to you and you're upset about your reviews, you know what to do.
0
0
5
This Kaleidoscopic Space Sculpting paper reminds me of the Infinity Mirror Rooms installation by Yayoi Kusama at Tate Modern.
A kaleidoscope consists of a camera looking at multiple mirrors. These mirror reflections can be used to sculpt complex 3D shapes.
1
1
5
@3DVconf
Hi, I've downloaded the author kit, but I don't see my paper ready yet. Is that a technical issue 🐞?
1
0
4
We can also sample new animal instances from the learned base shape bank and smoothly interpolate between various animal species.
1
0
4
Also thanks to other organizers for putting this together:
@chrirupp
@NataliaNeverova
Paolo & Andrea!
Hope it has been a rewarding full-day experience for all the attendees! Enjoy the rest of
@ICCV_2021
!
0
0
4
"Blocks-World Cameras" by Jongho Lee, Mohit Gupta
Probing block scene geometry with a pattern projector (I'm sure I've seen some similar ideas, let me know if you find them!)
1
1
4
"Intentonomy: a Dataset and Study towards Human Intent Understanding" by
@KMnP_menglin
et al.
A very interesting study on human "intent" in photos!
1
0
4
@dimadamen
@kaihan_cs
@BristolUniEng
@Oxford_VGG
Congrats
@kaihan_cs
!!! All the best for the exciting journey ahead!
0
0
4
#CVPR2020
Are there any instructions for authors during Q&A sessions?
Do we have control over the assigned Zoom rooms? How can we prevent it from being hijacked?
@CVPR
I've seen that happen at a workshop...
I'm sure
@CSProfKGD
knows as well :)
2
0
4