Roni Sengupta
@SenguptRoni
Followers
2,171
Following
870
Media
69
Statuses
743
Asst. Professor @unccs , Prev. postdoc @uwcse (UW), Ph.D. @umiacs (Univ. of Maryland). Computer Vision & Graphics (she/her)
Chapel Hill, NC
Joined April 2018
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Kroos
• 428521 Tweets
England
• 132219 Tweets
Reich
• 122861 Tweets
Boeing
• 61575 Tweets
#2024해피수호데이
• 60859 Tweets
#お迎え渋谷くん
• 52112 Tweets
招待コード
• 50234 Tweets
ソフトバンク
• 47675 Tweets
Rashford
• 46413 Tweets
Shadow of the Erdtree
• 45284 Tweets
Singapore Airlines
• 43541 Tweets
Modric
• 38373 Tweets
#メンテの日なのでssを貼る
• 33295 Tweets
Southgate
• 32690 Tweets
Starc
• 30521 Tweets
#KKRvsSRH
• 29158 Tweets
ROCKSTAR PRINCE SUHO DAY
• 28768 Tweets
Grealish
• 27690 Tweets
كروس
• 26516 Tweets
#日野下花帆生誕祭2024
• 25261 Tweets
GENTO AYAW PATALO
• 21501 Tweets
Iniesta
• 21458 Tweets
クロース
• 19554 Tweets
Shaw
• 18362 Tweets
花帆ちゃん
• 17155 Tweets
Hellblade 2
• 16743 Tweets
Vivian
• 15875 Tweets
Sancho
• 15807 Tweets
Tripathi
• 10506 Tweets
Pinned Tweet
Post
#CVPR2024
Holiday group photo with the lab and it’s ‘emotional support’ members!
Fall in North Carolina is so pretty. Come join us and have a fun grad school experience ☺️ we will do cool work in Computer Vision & Graphics!
5
4
100
Excited to release v2.0 of our Background Matting project, which is now REAL-TIME & BETTER quality: 60fps at FHD and 30fps at 4K! You can use this with Zoom, check out our demo! 👇
Webpage:
Video:
More in the thread! [1/6]
22
162
998
Updates on "Background Matting: The World is Your Green Screen", CVPR2020.
We have released our captured dataset and training code: .
Check out the amazing Google Colab created by Andrey for easy background replacement on your data: .
9
206
838
Stuck at home? Can't travel to take cool photos? Bored with video conferencing?
We bring you "Background Matting: The World is Your Green Screen" (CVPR 2020), which lets you replace your background without using any green screen studio.
Checkout:
6
93
446
Mandatory CVPR deadline photo with Pizza and Boba Tea! PI is behind the camera for moral support!
5
7
308
📢 I'm on the faculty job market📢
My goal is to create the next-gen of video communication. I use
#ComputerVision
&
#ComputerGraphics
to build equitable & accessible video production tools, requiring minimal human expertise, thus democratizing VFX.
Reach out if I'm a good fit!
11
43
274
We present "Shape and Material Capture at Home", which lets you recreate Photometric Stereo using household items -- walk around with a flashlight to reconstruct hi-res geometry and reflectance of an object.
(CVPR'21)
with
@daniel_lichy
Webpage:
[1/5]
7
53
228
Excited to share "A Light Stage on Every Desk". We learn a personalized relighting model by capturing a user watching a YouTube video.
Joint work with Brian Curless,
@kemelmi
, and Steve Seitz from
@uwcse
@UwRealityLab
Project Page:
[1/4]
7
39
168
Really happy to receive Best Student Paper Honorable Mentions at
@CVPR
2021! Congrats to two amazing undergrads
@peter9863
and
@andreyabtsev
at
@uwcse
. I am really fortunate to work with and mentor them. Cheers to the whole team :) Congrats to all other award recipients!
18
8
157
Our work has been accepted as Oral in CVPR 2021. Congrats to
@peter9863
and
@andreyabtsev
for leading this project!
Excited to release v2.0 of our Background Matting project, which is now REAL-TIME & BETTER quality: 60fps at FHD and 30fps at 4K! You can use this with Zoom, check out our demo! 👇
Webpage:
Video:
More in the thread! [1/6]
22
162
998
6
9
135
Background Removal a.k.a Alpha matting on videos by exploiting temporal information with a recurrent architecture. Does not require capturing background image or manual annotations.
Amazing results from
@peter9863
. Joint work with amazing folks from Bytedance (Linjie & Imran)
2
18
105
Checkout our latest work on Inverse Rendering of faces, accepted in CVPR 2018 (Spotlight).
From:
@akanazawa
@carlosdcastillo
David Jacobs (
@umdcs
@umiacs
)
Project Page (Code Published):
Arxiv:
0
34
91
"MVPSNet: Fast Generalizable Multi-view Photometric Stereo" lead by
@DongxuZhao2
to appear ICCV23.
Can we combine Multiview and Photometric Stereo to get the best of both worlds (global + local geometry) without NeRF style per-scene optimization?
Details:
0
19
88
Went to the UNC gym today and asked staff for help.
Me: "I am new here, can you please help me out?"
Staff: "Sure, are you a freshman?"
Me: "No I am just a new faculty!"
Wow, that kinda felt nice to be called a freshman! 😆
3
0
84
Really happy that our work, “Real-time High-resolution Background Matting” made the Best Paper candidates for
#CVPR2021
😃.
5
3
80
📢New research from our group
“Personalized Video Relighting with an At-Home Light Stage”
(1/3) We show how to leverage screen lighting as an 'at-home Light Stage' and develop a personalized relighting model. We can now replace your background and relight your faces to match it!
2
18
71
Thank you
@UNCResearch
for sharing my research and my personal journey of coming out as a trans woman 🏳️⚧️.
0
7
69
I miss those days where you come to work and have 1 thing to do: make your research work!
This faculty app season, I'm all over the place: writing statements, navigating portals, paper deadlines, advising students, brainstorming ideas, while being social!
I need a task manager!
5
0
66
📢 Our new work in CVPR'22.
Daniel continued his past work () of democratizing Photometric Stereo.
This shows that recursive neural nets are very efficient and produce HQ results in high-res.
Near-field lighting provides extra cues but is more challenging.
Pleased to announce our paper “Fast Light-Weight Near-Field Photometric Stereo” has been accepted to CVPR 2022
Reconstruct high quality geometry from multiple lights:
30x faster than SOTA
30% less GPU memory
13% better accuracy on LUCES dataset
Webpage:
1
23
92
1
7
63
When
#CVPR
is all about reunion and farewell…
The amazing crew from
@uwcse
GRAIL
@UwRealityLab
- the past, present, and the future! Will miss them all!
#CVPR2022
1
2
50
📢 New work from our group, exploiting the power of near-field illumination to improve monocular depth estimation of foundation model (DPT-Hybrid).
We can now predict accurate depth maps of internal organs from clinical endoscopy videos that will lead to better surgical outcomes
Challenges remain in 3D understanding from videos which lack strong geometric features and have challenging lighting conditions. Check out PPSNet - our monocular depth estimation (MDE) solution that achieves SOTA on endoscopy videos!
Project Page:
🧵
1
6
17
1
5
48
Looking forward to this Homecoming 🎉
We are excited to have
@SenguptRoni
visiting
@umdcs
(go Terp 🐢) later this month!
She will be giving an exciting talk on building personalized and efficient 3D models!
2
6
57
0
5
48
“A Light Stage on Every Desk” was accepted at
@ICCV_2021
!
Excited to share "A Light Stage on Every Desk". We learn a personalized relighting model by capturing a user watching a YouTube video.
Joint work with Brian Curless,
@kemelmi
, and Steve Seitz from
@uwcse
@UwRealityLab
Project Page:
[1/4]
7
39
168
1
3
44
I am so thankful to
@UNC
for having wellness days during the semester, not becoz it coincides with Valentines day, but becoz I can finally relax about not being ready for my class and try to write a proposal! It turns out junior faculties need wellness days more than students!
3
1
45
This is me when I am teaching Diffusion models in 2 weeks!
2
0
44
Done with all 6 CVPR reviews. I still have 1 more day to revisit them and make sure my summary is readable to the ACs! I was lucky enough to get all 6 papers on the same topic! Often, they are all over the place! Anyway, I hope I did a decent service to the community!
1
1
44
Introducing ‘Measured Albedo in the Wild’ (MAW), a new benchmark for evaluating albedo estimation in the wild (ICCP 2023).
MAW measures physically accurate albedo and evaluates using intensity, chromaticity, and texture metrics, vs relative judgement used in defacto WHDR metric.
1
11
42
Zoom & other video conferencing apps use segmentation, we estimate the alpha matte, just like a green-screen. This preserves hair level details when replacing the background.
[2/6]
1
2
41
Come join me and Daniel! We will be presenting our work at poster 193 tomorrow (Thursday) morning 10.00-12.30!
Pleased to announce our paper “Fast Light-Weight Near-Field Photometric Stereo” has been accepted to CVPR 2022
Reconstruct high quality geometry from multiple lights:
30x faster than SOTA
30% less GPU memory
13% better accuracy on LUCES dataset
Webpage:
1
23
92
0
4
41
Helpful advice from Jia-Bin!
One thing I found immensely helpful is to have a few friends/colleagues who are also in the faculty market to share the journey and the struggles :) Isolation is not at all helpful + you learn from each other!
How to get a tenure-track faculty job?
Interested in pursuing an academic career? Summer is the best time to kick off your search!
Here are some tips that I learned. 🧵
4
85
472
1
1
35
I did send this video to my slack group motivating my students to work hard for the CVPR deadline, saying 'this could be us'! Let's see how many papers end up getting accepted, if at all :P
Assistant professor traveling to conference with their first and only student
42
415
5K
0
0
34
OMG
@satyanadella
discussed our work on Background Matting and how Microsoft used it in their Virtual Build Conference, in his CVPR2020 Keynote talk. Check it out here:
with
@vivjay30
@kemelmi
#CVPR2020
0
5
35
To achieve real-time & high-res, we first predict coarse results at low-res along with an error prediction map that highlights regions that need further refinement. Then we refine only selective patches at high-res.
[4/6]
1
2
32
Few favorite CVPR papers from session 2:
1/ "UP-DETR: Unsupervised Pre-training for Object Detection with Transformers".
Proposes unsupervised pre-training of transformers, which otherwise requires huge labeled training data, by simply trying to detect random query patches.
1
5
32
So happy to see that our recent paper "Background Matting: The World is Your Green Screen" (with
@vivjay30
, Brian, Steve, and
@kemelmi
) was used for developing this technology "Virtual Stage". Check out the cool video that describes the system :)
How do you appear on stage from the safety of your own home? Easy, background matting, deep neural networks, and an
@Azure
Kinect (or two). Learn how you can do it too.
#MicrosoftAI
#MSBuild
4
92
200
3
5
31
Our Aging project got accepted in ECCV2020.
@royorel
will be soon releasing the code.
I’m proud to share that “Lifespan Age Transformation Synthesis” was accepted to
#ECCV2020
Our method synthesizes continuous aging for ages 0-70. Check out the project page and stay tuned for the code release.
@SenguptaUmd
@ohadf
@elishechtman
@kemelmi
0
8
45
0
3
30
Being on the job market and interacting with so many amazing folks taught me a lot! This was both the most exciting and stressful time of my life so far!
Can't believe that at one point during my Ph.D., I almost thought of quitting and going back to India!
2/3
1
1
30
Presenting "GaNI: Global and Near Field Illumination Aware Neural Inverse Rendering" () that reconstructs shape+BRDF of a scene captured with a co-located light & camera (flashlight on a phone).
👇 Results: re-rendering in novel views and lighting
🧵
1
5
31
Had a super fun conversation with
@PhilEdwardsInc
from
@voxdotcom
about segmentation vs matting and the future of AI green-screening.
0
6
27
Thanks a ton to David Jacobs
@davwiljac
(I am so lucky to have you in my life 🥹),
@akanazawa
, Steve Seitz,
@kemelmi
, Brian Curless,
@jbhuang0604
,
@CSProfKGD
,
@DBiswadip
, and many more folks! Thanks to my wife
@scentomentalgal
for her amazing support!
2
1
27
My life keeps swinging between 2 Capitol Hills, at the two ends of the country!
Also Happy Thanksgiving and amazed by the photo quality of the new iPhone13!
0
0
26
📢New research from our group ...
Can we personalize a 3D generative model using only a few images of an individual to better reconstruct, synthesize, and edit their 3D faces?
Yes, we can, using only 240K (<1 MB) separate parameters for each individual.
👇
Excited to share “My3DGen: A Scalable Personalized 3D Generative Model”. We learned a personalized 3D face generative model by finetuning a pre-trained 3D-GAN using only 50 selfies of an individual.
Project page:
1
6
26
0
6
26
Some more results! Our method works surprisingly well on challenging scenes with moving hair!
Credit goes to amazing
@uwcse
MS students
@peter9863
and
@andreyabtsev
(hire them!) for jointly leading the project and
@kemelmi
Steve Seitz & Brian Curless for their mentorship!
[6/6]
1
0
26
Thanks
@jbhuang0604
for inviting — had a fun visit at my alma talking about what we are working at UNC! 💗
Thanks
@SenguptRoni
for giving such an interesting talk on personalized and efficient 3D models!
Go Terp! 🐢
0
0
18
0
0
26
This is starting to look legit now!
Rare footage from my first year as a TT assistant professor 👀
50
532
6K
2
0
25
My fav CVPR paper series!
Session 1.
I have often wondered why VGG for style loss? Why not powerful ResNets? This paper analyses why and shows a simple softmax activation (named SWAG 😆) can greatly improve stylization performance.
4
0
25
Will be presenting our work in 30mins at poster 2A (8am PDT/11am EDT). Come and chat with me as I sip my morning coffee!
Excited to share "A Light Stage on Every Desk". We learn a personalized relighting model by capturing a user watching a YouTube video.
Joint work with Brian Curless,
@kemelmi
, and Steve Seitz from
@uwcse
@UwRealityLab
Project Page:
[1/4]
7
39
168
1
0
25
Our method even outperforms chroma-keying (green-screen) software on an amateur setup, since it's more tolerant to imperfect green-screen setups. Content creators can try out our method instead of setting up a green-screen at home!
[3/6]
3
1
24
😂😂 this is especially true for the greencard letters and profile that the lawyers create 😂
When the professor says “go ahead and write the rec letter, I’ll sign it.”
37
1K
9K
0
0
22
In comparison to our Background Matting v1.0 (BGM CVPR'20), which operates at 512x512 at 8fps, our current approach (v2.0) achieves better quality and is also faster (30fps at 4K)!
Check out our GitHub repo:
[5/6]
1
1
22
Come work with us! We are a very friendly department with lots of focus on high-quality research! The Triangle area is really nice to live in! (Didn't expect that before I moved in here!)
APPLY: TENURE-TRACK/TENURE FACULTY POSITIONS! Research areas include but not limited to: vision, systems, comp bio, security, medical image computing, data science, robotics & AR/VR. Join our community committed to research & teaching, diversity & collab.:
0
12
11
0
5
21
Interested in UNC-Intel Summer REU (for UG students at US institutions) working with me on developing cutting-edge Computer Vision algorithms for Deep Fake detection?
Deadline: Feb 25.
Feel free to DM me about this.
0
7
20
Super cool work that I enjoyed being a part of!
Check it out ⬇️
TLDR: Our paper proposes a "universal" capability for image generation that can use any off-the-shelf model trained on clean images to guide the diffusion process.
This eliminates the need for training and dataset requirements. However, it comes with the price of longer steps.
1
17
87
0
0
19
@jbhuang0604
I tend to focus more heavily on basics on 3d fundamentals, image processing operators, neural network based recognition. I feel these things are still relevant, and teaches them good math in linear algebra, calculus, probability.
2
0
20
Few fav papers from Session 3 & 4:
1/ "Omnimatte: Associating Objects and Their Effects in Video"
@erika_lu_
@talidekel
& team from
@Oxford_VGG
Decomposes a video into multiple foregrounds and a background layer with self-supervised learning, which can then be used for editing.
1
0
19
P.S. I messed up the youtube video link!! Here it is again:
3
0
19
You hear so much about active shooting and it feels distant until it happens to a building next to you! I hope the situation gets resolved soon! Stay strong UNC!
!AlertCarolina! Emergency–Update: Stay sheltered in place until all clear given. All classes & events cancelled for rest of day. Check for updates.
11
192
330
0
0
19
Happy CVPR rebuttal week everyone 🤣
0
0
18
Thanks
@CSProfKGD
for hosting me! It was fun and thanks to all the wonderful folks at Samsung AI Toronto for the insightful comments 😊
BIG thank you to
@SenguptRoni
for virtually dropping by today at the Samsung AI Centre Toronto and sharing your work with us! GREAT talk!
0
1
15
0
0
17
On a flight after 15 long months! Guess where I am going? ☺️
#MemorialDayWeekend
#BackToNormal
3
0
18
Come and work with us! I truly enjoyed the first 1.5yrs at UNC. The environment is very supportive for junior faculties and with lots of collaboration potentials across the board! And Chapel Hill-Carrboro is truly beautiful and eclectic!
APPLY: TENURE-TRACK/TENURE/TEACHING FACULTY. Research areas include but not limited to: ML, NLP, vision, graphics, systems, bioinformatics, security, medical imaging, robotics & AR/VR. Join our team - committed to research, teaching, and collaboration.
➡️
0
9
20
1
2
17
@jbhuang0604
Yes! very harmful suggestions. I have seen students who aim so high that every project is too low for them often blocking any progress. I keep telling them you can't go high without building one block at a time!
0
0
14
A perfect image to capture my obsession with alpha matte of hair 😂
1
0
17
Today, we will be presenting our work "Shape and Material Capture at Home" at Paper Session
#5
, Tuesday, June 22 (10pm-11.59pm EDT, 7pm-9pm PDT).
Come and say Hi!
We present "Shape and Material Capture at Home", which lets you recreate Photometric Stereo using household items -- walk around with a flashlight to reconstruct hi-res geometry and reflectance of an object.
(CVPR'21)
with
@daniel_lichy
Webpage:
[1/5]
7
53
228
1
1
17
4/ Super impressive result "Robust Consistent Video Depth Estimation"
@JPKopf
@jastarex
@jbhuang0604
0
5
16
Check out our new work "Lifespan Age Transformation Synthesis" with Roy Or-El,
@SenguptaUmd
,
@ohadf
,
@elishechtman
, and
@kemelmi
.
Project Page:
3
1
15
@CSProfKGD
A lot of vision and ML folks have electrical engg background! For me i wanted to study signal processing as an EE major, but images turned out to be more visually appealing 😂
2
0
15
I had a call with my advisee
@daniel_lichy
about some technical problems in 3D. I called 'Daniel' from my phone and we talked for 5 mins about this before being told he is not the same 'Daniel'. He was also expecting a call to talk about some 3D issue, which sounded similar!
0
0
15
Wow! Learning to play soccer!
From Motor Control to Team Play in Simulated Humanoid Football
pdf:
abs:
video:
6
253
994
0
0
14
Now you can try our background replacement with this web app from
@GradioML
!
We built a demo of Real-Time High-Resolution Background Matting 🚀
Awesome work by
@SenguptaUmd
,
@peter9863
et al.
Demo:
Paper:
Repo:
1
4
12
0
2
14
Unfortunately, neither the lead author
@DongxuZhao2
nor I can attend
#ICCV2023
due to visa issues ☹️
However, Jan-Michael Frahm will be there presenting the poster tomorrow, Thursday 2.30-4.30pm, in Room "Nord" - Poster 043!
"MVPSNet: Fast Generalizable Multi-view Photometric Stereo" lead by
@DongxuZhao2
to appear ICCV23.
Can we combine Multiview and Photometric Stereo to get the best of both worlds (global + local geometry) without NeRF style per-scene optimization?
Details:
0
19
88
0
0
14
Cool application of our Background Matting work. Cheers to Policarpo for creating this Quarantine compilation.
Policarpo Del Canto Baquera, a Computational Design graduate student in
@CMUSoA
, depicted
@Quarantine
life in 360° with the help of ML-powered rotoscoping:
2
8
54
1
2
12
@CSProfKGD
You can start a crash course/blogpost for young faculties by sharing your tips and strategies on teaching, especially in Comp. Vision.
2
0
13
I know everyone is super stressed with the election results and the CVPR deadline, but here is something funny that a friend of mine made with my Halloween costume.😆
#DontStopTheCount
#STOPTHECOUNT
#ElectionResults2020
#CountEveryVote
#CountDracula
1
1
13
Excellent Talk by
@sellan_s
and
@ana_dodik
on gender in CG research. Planning to include this in my lectures. Separate lectures on ethics and bias in a Vision or Graphics class is becoming increasingly important (see
@Jimantha
's Computer Vision course).
Our talk (
@ana_dodik
's and I) on gender in computer graphics in research is finally on Youtube. We worked pretty hard on this! Hopefully enough of you will watch it that it'll start a conversation in our field:
0
17
62
1
2
13
Shout out to
@daryl_imagineai
for creating an android app for Background Matting. Try it out and change your social media photos without actually traveling 😝. Just take an additional image of the background without you and then you can use any background!
2
1
13
I was making my lecture slides for my Computer Vision course for the 1st time! It's been 10years since I listened to 'The Fundamental Matrix' song! The nerdiest and the best part of tomorrow's lecture! I have never met Daniel Wedge, but you are a legend!
0
1
12
This is super cool and creative!
Some of my early experiments with
@runwayml
GEN-1
It’s still early days of AI video - this tech will only get better. A whole new generation of filmmakers is gonna be able to make whatever they want on zero budget
See below for my process
127
1K
8K
0
0
12
Absolutely golden suggestions! Also, I really do enjoy the process of making good videos, websites, github and twitter posts! So much fun in reaching out to a broader audience!
Sharing ideas on how to disseminate your research.
"I am THRILLED to share that our paper is accepted to ..."
Congrats! So what's next? No one is going to browse through the list of thousands of accepted papers. Ain't nobody got time for that.
Check out 🧵below for examples.
11
181
732
0
2
11
Thesis defended!! Thanks to all the members of my committee and especially my advisor Prof. David Jacobs. Next stop: Postdoc at
@UwRealityLab
.
1
0
11
Curious to know how you or your kids will age? Try out a quick Colab demo of our paper "Lifespan Age Transformation Synthesis" (ECCV'20).
I look pretty cute as a kid and cool as a grandpa! Feel free to share your cool transformation videos...
Colab:
I had an "I had a joke" joke, but it wore out quickly and didn't age well.
You know what does age well? Our newly released "Lifespan Age Transformation Synthesis" code !!!
Code:
Colab:
Paper:
0
1
5
2
1
11
Seriously? 🤔 I guess my parents did not know that I would be an academic and gave me a formal name that they don’t even care to pronounce ever 😂 Henceforth I will just use my nickname ‘Roni’ (yes that’s what my parents call me!).
having a name that takes longer to pronounce is associated with a significantly lower likelihood of being placed into an academic job or obtaining a tenure track position & an initial placement at an institution with lower research productivity; econ
61
226
980
0
0
11
3DV social media page is providing me with so many awesome memes for my lectures!
1
0
11
So excited to see so many amazingly talented new folks joining UNC CS
@unccs
!!
🎉Please join us in welcoming this year's new faculty cohort! From algorithms, security, machine learning, to graphics and computational optics, these faculty will maintain the standard of excellence at
@UNCCS
!
➡️
@UNC
@unccollege
@UNCResearch
@UNCSDSS
4
9
39
0
1
9
@KostasPenn
mostly naive optimism ... I have seen people who are not sure of submitting 2 days before the deadline, waiting for divine interventions 🤣
Also there are some curious folks who just wants to know the latest paper ID few hours before the deadline, creating fake submissions 🤣
1
0
10
Come work with us at UNC!
Super excited to be in a very friendly and young CS department! Extremely happy with so many talented UGs that have started working in my lab!
UNC CS is hiring for multiple Teaching Assistant Professor positions! Come work with amazing undergrads and a faculty enthusiastic about teaching CS in an inclusive, creative, and diverse community thanks to
@unccs
being rooted in our liberal arts college!
0
9
23
0
0
10
Our setup only requires a $5 flashlight and a sweet potato (in absence of a tripod). No need for darkroom or light stage capture!
Credit goes to
@umdcs
Ph.D. candidate
@daniel_lichy
for leading the project, along with Jiaye Wu and David Jacobs (
@davwiljac
) for mentorship.
0
0
9
It was a pleasure chatting with you
@Lialialiacole
, and thanks
@runwayml
for creating this platform to bring young AI researchers and artists together for conversations. I got to learn a lot from Lia :)
Thank you for chatting
@SenguptRoni
😊 I found it especially inspiring how he chooses chooses what to research-
😒Less: "With a depth camera & this other fancy equipment, we can do X"
✅🤳More: "What can we do w the photos in your phone right now?"
0
0
11
0
1
10
@tomgoldsteincs
There are many great papers that were rejected on the first attempt but being on social media provided great visibility and helped other researchers to take inspiration and built on it.
0
0
10
We will be live tomorrow 12-2 PDT (both am and pm) for CVPR poster session. Come chat with us at Poster
#104
.
Our CVPR talk is now live for Background Matting The World is Your Green Screen! We'll be presenting during poster session 1.2 from 12-2pm PT (Tuesday) and again 12 hours later. Come check it out! With
@SenguptaUmd
@kemelmi
0
6
28
0
1
10
'Class-related Memes' is going to be a 5 point assignment in my Fall 2023 Computer Vision class! Meme creativity will lead to better paper writing! If you don't believe ask
@jbhuang0604
🤣
As part of a final assignment in my seniors biophysical chem class, I asked students to share a class-related memes, with ones they make themselves highly encouraged!
For everyone's enjoyment, here are some of my favorites!
#chemistry
#biochemistry
26
132
2K
0
0
9
Your paper/project release is never complete without a tweet from
@ak92501
;)
Thanks for sharing all the beautiful works, every day!! What a great service you are doing to the community! 👏👏
Real-Time High-Resolution Background Matting
pdf:
abs:
project page:
github:
0
23
178
0
0
10
@tomgoldsteincs
Lots of important research discoveries are made by accident! A coding bug is the CS version of that accident I guess?
1
1
9
I have been collaborating with Xin for a while! He would be a great hire!
0
0
9
Super cool work! These kinds of fun and cool work attracted me to work on Computer Vision and Graphics, and I am glad that the community is still producing plenty such works!
Can you make a jigsaw puzzle with two different solutions? Or an image that changes appearance when flipped?
We can do that, and a lot more, by using diffusion models to generate optical illusions!
Continue reading for more illusions and method details 🧵
15
119
602
1
0
9
Aging Monalisa ...
Try out your own aging results:
0
0
8
Pretty cool result in synthesizing fluid motion in natural images!
Excited to show off our new project on single-image cinemagraphs. Our method automatically turns a _single image_ into a seamlessly looping video!
Website:
Video:
w/ Brian Curless, Steve Seitz, Rick Szeliski
More in thread! [1/5]
30
386
2K
0
0
8
A perfect scene from today's game: Let's view this nutmeg goal in 3D or 6DoF view synthesis! Nutmegs are an absolute delight to watch, especially in 3D!
1
0
8