Incoming robotics faculty @JHUCompSci | sci-fi-lab.github.io/
Berkeley, CA
- Tweets 1,294
- Following 2,036
- Followers 5,140
- Likes 6,052
Pinned Tweet
In Fall 2026, I will begin a tenure-track faculty position @JHUCompSci
Announcing the SciPhy lab, where we will study the science of physical agents (robots)
We are now recruiting our first cohort of PhD students. If this is you, see
sci-fi-lab.github.io/
36
78
482
31,353
Today we release gradslam - automagically differentiable SLAM. Run dense 3D reconstruction in @PyTorch!
And, it's fully differentiable :) Co-led w/ @S_Saryazdi, ably supported by @mautkiungli @duckietown_coo
Webage: gradslam.github.io
Paper: arxiv.org/abs/1910.10672
21
253
1,297
Announcing our newest preprint
Features from CLIP, DINO are ready for zero-shot 3D perception.
𝗖𝗼𝗻𝗰𝗲𝗽𝘁𝗙𝘂𝘀𝗶𝗼𝗻 builds open-set multimodal 3D maps by fusing features to 3D. These maps can be queried by text, image, click, and audio📜👇
concept-fusion.github.io/
10
124
584
154,478
Role update: after 4 wonderful years @MontrealRobots @Mila_Quebec , I’ll move to MIT in early 2022 for a postdoc w/ Josh Tenenbaum and Antonio Torralba
(This is all once pass my thesis defence ofc :) )
Their groups have an amazing set of colleagues I look fwd to working with!
Welcome Ayush Tewari (@_atewari) and Krishna Murthy (@_krishna_murthy), who have joined MIT as post-docs! Thrilled to have you as colleagues :) Lots more cool work on Neural Rendering, Scene Representations, and ML for Physics to come :D
34
8
353
🚨 ConceptGraphs: Open-Vocabulary 3D Scene Graphs for Perception and Planning
RGB-D images -> 3D scene-graphs -> enhance with vision-language features
Enables a wide range of perception and planning abilities. Deployed in many real-world use-cases.
concept-graphs.github.io/
11
69
340
76,574
Excited to announce our new work!
"gradSLAM: Dense SLAM meets automatic differentiation"
We leverage the power of autodiff frameworks to make dense SLAM fully differentiable.
Paper: arxiv.org/abs/1910.10672
Project page: montrealrobotics.ca/gradSLAM…
Video: piped.video/watch?v=2ygtSJTm…
10
85
269
Delighted to share that I recently defended my PhD dissertation
𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁𝗶𝗮𝗯𝗹𝗲 𝗪𝗼𝗿𝗹𝗱 𝗣𝗿𝗼𝗴𝗿𝗮𝗺𝘀
krrish94.github.io/files/Jat…
Extremely grateful to @duckietown_coo @MontrealRobots and @Mila_Quebec for their wonderful support over the last four years!
28
8
270
Honored and thrilled to have received an @NVIDIAAI PhD fellowship!! blogs.nvidia.com/blog/2020/1…
I'd like to acknowledge the support of my advisor @duckietown_coo and close mentors @DerekRenderling @FidlerSanja who make work so much fun :) @UMontrealDIRO @Mila_Quebec @MontrealRobots
30
5
229
I've ported the (excellent!) Neural Radiance Fields (NeRF) paper to @PyTorch. Try the (tiny-NeRF) Colab notebook at colab.research.google.com/dr… or get the full code at github.com/krrish94/nerf-pyt…
NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
Input is images with known camera poses. Can't wait to disable all the sanity checks and see what this renders if I give it impossible geometry.
abs: arxiv.org/abs/2003.08934
site: matthewtancik.com/nerf
1
59
223
Gradschool applicants from India: I know a large fraction of profs at Indian universities ask you to draft your own letters of recommendation. This is quite unfortunate! (1/6)
2
26
221
At @iclr_conf, we're excited to present 𝗴𝗿𝗮𝗱𝗦𝗶𝗺, a differentiable simulator 👇
gradSim enables visuomotor control and physical parameter estimation directly from images/videos
arxiv.org/abs/2104.02646
Co-led with @milesmacklin
gradsim.github.io/
2
36
211
At @ICCV_2021, we are hosting a workshop on "Differentiable 3D vision and graphics" with a broad focus on differentiable programming and inductive biases for vision/graphics.
Submissions are now open! Visit montrealrobotics.ca/diff3d/ for more details
1
27
169
Student pricing for virtual conferences:
𝗠𝗟
ICLR: $25 (iirc)
ICML: $25
𝗥𝗼𝗯𝗼𝘁𝗶𝗰𝘀
ICRA: €25
RSS: $20
𝗚𝗿𝗮𝗽𝗵𝗶𝗰𝘀
Eurographics: Free
SGP: Free
HPG: $15
Siggraph: $125
𝗖𝗩
CVPR: $125
ECCV: £450 (for authors)!
@amlankar95 @adityakusupati
This is the last @eccvconf paper I will submit if they continue with Springer. 450 GBP publication fee is more than the price of 20 ICML student registrations.
#ECCV2020 #OpenAccess
5
23
157
Robot Learning @JHUCompSci is growing!
Multiple faculty incl. @mangahomanga @zihyunchiu @HaiminHu @jmin__cho @anand_bhattad and these are all just ~new hires with @HopkinsDSAI
Apply to work with us this PhD admissions cycle (Dec 15 deadline)
cs.jhu.edu/academic-programs…
4
33
162
16,042
Announcing the robot learning seminar series at @MILAMontreal and @MontrealRobots. Great speaker lineup! See montrealrobotics.ca/robotlea… for schedule. First talk this Friday (i.e., tomorrow/today) by @florian_shkurti!
2
40
148
Thanks @iclr_conf for listing me an "outstanding reviewer"!
TBH, my assignments were spot-on.
I wish the free registration were transferrable. I'm fortunate to be in a well-funded program; transferring the ticket might enable someone less privileged attend.
4
2
151
Excellent advice to early career researchers (phd / postdocs) at the RSS pioneers workshop today (Dieter Fox, @ken_goldberg, Jana Tumova, Sangbae Kim). Shoutout to the organizers @valentinp @p_alvesoliveira Kim Baraka, Noha Radwan, and others. @RoboticsSciSys @rssconf #RSS2020
4
32
139
Announcing our @NeurIPSConf workshop titled "Physical Reasoning and Inductive Biases for the Real World".
What forms of inductive bias (if any) do we need for physical reasoning in the real world?
Submissions due 11 October
physical-reasoning.github.io…
2
20
127
Stoked to release work done as a summer intern at NVIDIA!
Kaolin: A Pytorch library for Accelerating 3D DL research
Paper: arxiv.org/abs/1911.05063
Code: github.com/NVIDIAGameWorks/k…
Work w/ @EdwardJamesSmi4, JF, @Caenorst , Artem, Tommy, @RevLebaredian @gavrielstate @FidlerSanja
1
29
130
The very NeRF we all admire just got 9x faster!
Grab my pytorch port here:
github.com/krrish94/nerf-pyt…
Same stunning visuals:
I've ported the (excellent!) Neural Radiance Fields (NeRF) paper to @PyTorch. Try the (tiny-NeRF) Colab notebook at colab.research.google.com/dr… or get the full code at github.com/krrish94/nerf-pyt…
6
18
121
Happening today!
"Differentiable 3D Vision and Graphics" workshop @ICCV_2021
Tune in to listen to an exciting speaker lineup - begins 1 pm EDT
Schedule: montrealrobotics.ca/diff3d/s…
ALT ICCV workshop on differentiable 3D vision and graphics features Yaron Lipman, Tzu-Mao Li, Nikhila Ravi, Wenzel Jakob, Katja Schwarz, Vincent Sitzmann, Clement Fuji-Tsang, Cengiz Oztireli, Marie-Julie Rakotosaona, and Matthew Tancik
1
19
106
gradSLAM is now an @icra2020 paper. More on this to come! Kudos to the anonymous reviewers for extremely detailed reviews.
@duckietown_coo @mautkiungli @MILAMontreal @UMontrealDIRO @CMU_Robotics
Excited to announce our new work!
"gradSLAM: Dense SLAM meets automatic differentiation"
We leverage the power of autodiff frameworks to make dense SLAM fully differentiable.
Paper: arxiv.org/abs/1910.10672
Project page: montrealrobotics.ca/gradSLAM…
Video: piped.video/watch?v=2ygtSJTm…
4
21
103
I am now in the Boston area, beginning my postdoc at @MITCoCoSci and @MIT_CSAIL
If you are in the area or are interested in differentiable/probabilistic programming, inductive biases for robot learning, I'd love to chat :)
Role update: after 4 wonderful years @MontrealRobots @Mila_Quebec , I’ll move to MIT in early 2022 for a postdoc w/ Josh Tenenbaum and Antonio Torralba
(This is all once pass my thesis defence ofc :) )
Their groups have an amazing set of colleagues I look fwd to working with!
1
1
99
Six papers accepted to @ieee_ras_icra 2024 in Yokohoma 🧵
1. ConceptGraphs: Open-vocabulary 3D scene graphs for perception and planning
co-led w/ @alihkw_ @qiaogu1997 @SachMorin
🚨 ConceptGraphs: Open-Vocabulary 3D Scene Graphs for Perception and Planning
RGB-D images -> 3D scene-graphs -> enhance with vision-language features
Enables a wide range of perception and planning abilities. Deployed in many real-world use-cases.
concept-graphs.github.io/
3
12
102
11,846
We're hosting a Neurips 2020 workshop on "Differentiable computer vision, graphics, and physics in ML", with a fantastic set of panelists and speakers. Our call for short papers is now live. We hope to see your work! Visit montrealrobotics.ca/diffcvgp…
ALT Neurips 2020 workshop on differentiable computer vision, graphics, and physics in machine learning
3
20
99
Thank you @ieeeras for listing me an outstanding reviewer for RAL.
Credit to the associcate editors for nominating me, and for patiently chasing some of these reviews :)
ieee-ras.org/publications/ra…
ALT IEEE Robotics and Automation Letters Special Recognition Distinguished service as outstanding reviewer 1. Nouman Ali, Mirpur university of science and technology, Pakistan 2. Wolfgang Hoenig, Queensland University of Technology, Australia 3. Krishna Murthy Jatavallabhula, Universite de Montreal 4. Horia Porav, University of Oxford, UK
3
2
72
We're presenting gradSim today at @iclr_conf!
Join us at poster spot B5 in about 25 minutes from now
iclr.cc/virtual/2021/poster/…
At @iclr_conf, we're excited to present 𝗴𝗿𝗮𝗱𝗦𝗶𝗺, a differentiable simulator 👇
gradSim enables visuomotor control and physical parameter estimation directly from images/videos
arxiv.org/abs/2104.02646
Co-led with @milesmacklin
gradsim.github.io/
1
11
71
Happening tomorrow! #NeurIPS 2021 workshop on physical reasoning and inductive biases for the real world!
Join us at 11 am EST (8 am PST, 4 pm GMT/UTC)
For schedule and accepted papers, see physical-reasoning.github.io
Fantastic speakers and panelists from industry/academia
ALT Speakers/panelists at the workshop include: Tomer Ullman, Nils Thuerey, Karen Liu, Kelsey Allen, Kyle Cranmer, Shuran Song, Kenneth Tran, Masahiro Ono, Aleksandra Faust, Michael Roberts
2
15
65
Unsure about most images, but this one's definitely worth a thousand words!! A witty intuition for visual-inertial issues.
Whoever came up with this :)
2
10
65
To @ICCV_2021 area chairs looking for emergency reviewer(s), I'm more than happy to review a paper or two.
Over the past year, I've developed a deep respect for emergency reviewers who've come to the rescue at workshops I helped co-organize.
I hope to pay this forward :)
3
1
65
I'm honored to be selected to the RSS pioneers cohort of 2020, alongside 27 amazing PhD students and postdocs! piped.video/bbEqwyVtJXU @rssconf @MILAMontreal @UMontrealDIRO
2
2
60
We are demo-ing ConceptFusion at @CVPRConf for the next three hours or so.
Stop by the demo booths to chat with me, Ali, Qiao!
Announcing our newest preprint
Features from CLIP, DINO are ready for zero-shot 3D perception.
𝗖𝗼𝗻𝗰𝗲𝗽𝘁𝗙𝘂𝘀𝗶𝗼𝗻 builds open-set multimodal 3D maps by fusing features to 3D. These maps can be queried by text, image, click, and audio📜👇
concept-fusion.github.io/
7
61
10,824
Addendum: If you're a gradschool applicant and have been asked to "write a draft letter for yourself", here's my short summary on what makes a letter more informative
(Caveat: Assumes access to a lot of information only your referee might have)
krrish94.github.io/blog/grad…
3
7
59
New work, spearheaded by Nikhil and Jonathon, where we use 3D Gaussians as the map representation for dense SLAM.
Think Point/ElasticFusion, but with volumetric, opposed to surface, primitives.
Some very cool demos on the project page!
SplaTAM: Splat, Track & Map 3D Gaussians for Dense RGB-D SLAM
spla-tam.github.io
We extend Gaussian Splatting to solve SLAM, i.e., automatically calculate the camera poses when fitting the Gaussian scene from RGB-D videos.
Try it on your own iPhone capture today! 🧵👇
1
12
61
6,918
This @ICCV_2021 lot was probably the best set I've received in sometime. Thanks to the anonymous ACs for sending me wonderful reads, mostly :)
7 papers, ~30 hours of review effort; avg review length: 4.99 k chars, stddev: 1.1k chars
On to the discussion phase
2
2
57
Since it's grad school apps time -- I decided to finish up a series of articles i've stalled for long
To save myself the time to write the same replies/remarks all over again.
Will keep adding to this thread (1/n)
3
11
53
Jumping on to the bandwagon, following @michellearning @florian_shkurti @hoeferse
Posting 9 classic robotics/vision papers that inspire my work
I had a "summer of introspection" in the first year of my PhD: went back to read classics from the 80s and 90s (1/n)
1
2
49
It's always nice to be recognized for your reviewing efforts -- thanks for listing me among the set of "outstanding reviewers" @CVPR #CVPR 2021
cvpr2021.thecvf.com/node/184
1
1
47
If you're a workshop organizer looking for an easily customizable template for your event, consider using this: github.com/krrish94/sniffle-…
Nothing fancy -- some Jekyll collections to automatically curate speaker/organizer/schedule/proceedings pages.
krrish94.github.io/sniffle-w…
3
46
Honoured to have won the best RAL paper award, announced at @icra2020 earlier today, for our paper: "MapLite" krrish94.github.io/publicati…
Teddy did a fantastic job leading the work! Check it out on ICRA Slack #tub06_3, if you're attending.
Very excited to have received a Robotics and Automation Letters Best Paper Award with Teddy Ort, Krishna Murthy, @rohanbbanerjee, @saikrishna_gvs, @dhaivat1729, @duckietown_coo, and Daniela Rus for MapLite (ieeexplore.ieee.org/document…). Huge shout out to Teddy for leading that work!
2
4
45
Our work "gradSLAM: Dense SLAM meets automatic differentiation" will be presented at this @CVPR workshop tomorrow, among a great speaker lineup! Joint work: @mautkiungli Soroush Saryazdi @duckietown_coo #CVPR2020 montrealrobotics.ca/gradSLAM…
Sunday at #CVPR2020: Our workshop on Deep Declarative Networks with invited talks by @brandondamos, @RickyTQChen, @chelseabfinn, @FuaPv, @MajiSubhransu and @zicokolter, and paper presentations from @ZGojcic, @egrefen, @krrish94 and more.
Full schedule: cvpr2020.deepdeclarativenetw…
3
46
I am in New Orleans this week, for CVPR. I’d love to catch up with folks in-person after a while.
DM / email if you’d like to meet
1
42
A very interesting (and comprehensive!) proposition from @lucacarlone1's group.
3D Dynamic Scene Graphs: Actionable Spatial Perception with Places, Obje... piped.video/SWbofjhyPzI via @YouTube
17
43
New work led by Dhaivat. A simple trick to calibrate uncertainty estimates from neural nets.
Excited to share our new work:
"f-Cal: Calibrated aleatoric uncertainty estimation from neural networks for robot perception"
A method to calibrate aleatoric uncertainty using first principles of calibration.
Preprint: bit.ly/39OeNIC
Webpage: f-cal.github.io
1/6
5
42
Participate in our @iclr_conf initiative to "rethink" whether static PDFs are the best medium for disseminating research w/ @khimya @adityakusupati @breandan Bhairav @amy_tabb @tegan_maharaj @annargrs @sarahookr @deviparikh @PoloChau @DerekRenderling Yoshua @Mila_Quebec
1/N We are pleased to introduce our @iclr_conf 2021 workshop on - Beyond static papers: rethinking how we share scientific understanding in ML. We invite discussion on whether static PDFs are the best format for scientific papers going forward. rethinkingmlpapers.github.io…
1
11
41
CVPR reviewer update:
- uploaded reviews on 5/6 papers (1 withdrew)
- one enthusiastic AC already initiated discussions on papers
- upped my rating on 2 papers (authors: rebuttals work!)
- trying to save 2 papers (fellow reviewers: bat for papers you enjoyed reading!)
#CVPR2021
1
39
Same boat! Reviewing ML PhD applications. I'd probably not have secured an admit if I were to compete with this peer group!
Many applicants with great publication record, stellar letters, grades! I have a feeling advisor/school fit is going to be the deciding factor here.
Reviewing PhD applications this year in ML/AI. At this point we might as well just say applicants already need a PhD to get accepted.
Is this the case with every field, or is it just ML/AI which has gotten so competitive? Thank God I applied 4 years ago!
2
1
38
A very fun collaboration with @Nik__V__ Avneesh Jay @sourav_garg_ and others!
We deployed our place rec system indoors, outdoors, aerial, underwater, subterranean envs.
Use this as a drop-in to your next localization/SLAM pipeline.
See Nikhil’s thread for more 👇
🧠🔬 Excited to share
AnyLoc: Towards Universal Visual Place Recognition
Foundation Models meet VPR
- VPR anywhere🌍🌊🏙️, anytime🌌☁️🌄, and under anyview🚡🚗🛸
- no retraining/finetuning 🔁
- aimed at general-purpose localization & navigation
anyloc.github.io/ 🧵👇
5
39
4,354
Join us this Friday at the Neurips workshop on Differentiable computer vision, geometry, and physics! Exciting speaker lineup!! Our schedule and accepted papers are up on our webpage montrealrobotics.ca/diffcvgp… @MontrealRobots @Mila_Quebec @UofTRobotics (1/3)
1
11
37
Here's a presentation @MandanaSamiei, Benjamin, and I did for the @DuckietownAI class, on what we believe SLAM currently lacks, and where it should go. Agree with @AjdDavison on many aspects. Aptly titled: "From SLAM to #SpatialAI"
docs.google.com/presentation…
1
11
37
Pointcloud classifiers are quite finicky when operating on data outside of the training distribution. This work from @autodesk (led by Saeid and Jieliang) introduces RobustPointSet: a new benchmark and evaluation protocol. arxiv.org/abs/2011.11572 github.com/AutodeskAILab/Rob…
2
6
33
Announcing new work:
Taskography - Evaluating robot task planning over large 3D scene graphs
Will be presented at CoRL today in posters & demos session 8
Webpage: taskography.github.io/
Video abstract: piped.video/mM4v5hP4LdA
1
6
36
The robot learning seminar series is back, starting this week!
We are organizing another edition of the robot learning seminar series.
montrealrobotics.ca/robotlea…
Talks usually at 4 pm ET
Subscribe to our channel bit.ly/2XvHuDT and get our calendar bit.ly/3i4od5N
@Mila_Quebec @UMontrealDIRO @NCRN_RCRC (1/2)
6
33
𝗧𝗵𝗿𝗲𝗮𝗱: I think conferences / academic events must continue to be predominantly online even post-pandemic.
6
2
33
New work out.
𝗠𝗼𝗻𝗼𝗟𝗮𝘆𝗼𝘂𝘁: 𝗔𝗺𝗼𝗱𝗮𝗹 𝘀𝗰𝗲𝗻𝗲 𝗹𝗮𝘆𝗼𝘂𝘁 𝗳𝗿𝗼𝗺 𝗮 𝘀𝗶𝗻𝗴𝗹𝗲 𝗶𝗺𝗮𝗴𝗲
Single-image -> Bird's-eye view @ 32 fps!
Work led by @HbutsuakK
projectpage: hbutsuak95.github.io/monolay…
arXiv: arxiv.org/abs/2002.08394
@wacv2020
1
10
33
This Friday at 4 pm ET, we have Ronald Clark @ronnieclark__ talking about "Representation learning for 3D vision". Watch the livestream on piped.video/watch?v=fLjDMKnm… @MontrealRobots @MILAMontreal
Announcing the robot learning seminar series at @MILAMontreal and @MontrealRobots. Great speaker lineup! See montrealrobotics.ca/robotlea… for schedule. First talk this Friday (i.e., tomorrow/today) by @florian_shkurti!
13
33
Lastly, to all referees that write their own letters - THANK YOU! Writing reference letters is a thankless job; and occupies a huge chunk of time (not to mention all the questionnaires to be filled out for each school-student pair). (6/6)
3
1
32
Check out @iclr_conf spotlight PAC-NeRF, which integrates a differentiable physics engine with a NeRF.
Estimates shape and physical properties of deformable, granular, and fluid media from multi-view videos.
Led by @xuanli917, a summer intern with the fantastic @gan_chuang
#ICLR23 Can NeRF understand physics? Please check out PAC-NeRF, which can estimate both the geometry and physical parameters of highly dynamic objects from multi-view videos!
Project page: sites.google.com/view/PAC-Ne…
Code: github.com/xuan-li/PAC-NeRF
6
30
3,815
"Thank you" to the anonymous CVPR AC who helped correct a grave typo in one of my emergency reviews. Need more people like you @CVPRConf
1
2
31
We are organizing a live poster session for our ICRA 2020 paper "gradSLAM: Dense SLAM meets automatic differentiation" tomorrow (June 01) at 11 AM EDT. Tune in to pick our brains, and see what's hot and brewing in the "differentiable SLAM" area. @mautkiungli @duckietown_coo
Excited to announce our new work!
"gradSLAM: Dense SLAM meets automatic differentiation"
We leverage the power of autodiff frameworks to make dense SLAM fully differentiable.
Paper: arxiv.org/abs/1910.10672
Project page: montrealrobotics.ca/gradSLAM…
Video: piped.video/watch?v=2ygtSJTm…
1
9
31
If you're a prof that requests students write their own letters - please reconsider. You'll do a great deal of good spending an hour or so drafting an informative letter. (5/6).
1
1
30
Our lab now has a Twitter handle :) Follow @MontrealRobots for more updates
3
30
Deadline for the Neurips 2020 workshop on "Differentiable computer vision, graphics, and physics in ML" is fast approaching (Oct 15). We look forward to seeing your submissions! @MILAMontreal @MontrealRobots @UofTRobotics
We're hosting a Neurips 2020 workshop on "Differentiable computer vision, graphics, and physics in ML", with a fantastic set of panelists and speakers. Our call for short papers is now live. We hope to see your work! Visit montrealrobotics.ca/diffcvgp…
ALT Neurips 2020 workshop on differentiable computer vision, graphics, and physics in machine learning
8
28
I will be @talkingrobotics next week. This will be about our recent work (co-led with @agiachris) on evaluating robot task planning over large 3D scene graphs
Our last session before Christmas 🎅 will be with Krishna @_krishna_murthy @Mila_Quebec @MontrealRobots about evaluating #robot task #planning over large 3D scene graphs. This work was just presented @corl_conf 👇
📅 Dec 10 | 9 AM PT - 6 PM CET
🔗 us02web.zoom.us/meeting/regi… #AI
3
28
Neural RGB->D. Cool results from folks at @NvidiaAI, CMU, @SenseTime_AI .
Paper: arxiv.org/abs/1901.02571
7
29
Achievement unlocked ;)
If you're working on anything releated to 3D deep learning, try Kaolin out, and tell us how we can improve it.
github.com/NVIDIAGameWorks/k…
1
29
Cool new work from Luca's group! Excited to see where the 3D scene graph space evolves to.
Big upgrade over existing techniques is the ability to build this map on-board in real-time.
See Luca's thread below for more
with SegmentAnything+CLIP, we can build rich metric-semantic 3D maps for robotics. but what is the right granularity for objects & places in the map? is a backpack an object? or is the zipper on the backpack an object?
piped.video/m-HJO10qhSQ
[1/n]
1
29
3,671
Great to have you talking about Kaolin! :) @AnimaAnandkumar
@FidlerSanja
@AnimaAnandkumar talking about Kaolin , @nvidia's Pytorch library to accelerate 3D vision, and other cool advances by Nvidia! Check out Kaolin here: news.developer.nvidia.com/ka…
Nvidia Clara for healthcare solutions:
developer.nvidia.com/clara
1
1
26
One thing I like about @iclr_conf this time around is ...
no more packed weekends with workshops to attend.
This is a welcome move and hopefully other conferences will follow suit
2
26
I am in NYC, attending RSS. Would love to meet other people at the conference. DM/email if you’d like that too
25
Next up in the Robot Learning Seminar Series at @MontrealRobots, we're glad to host Ankur Handa (@ankurhandos). Tune in at 4 PM EDT to watch the livestream piped.video/2BPa1YFhSL4
Announcing the robot learning seminar series at @MILAMontreal and @MontrealRobots. Great speaker lineup! See montrealrobotics.ca/robotlea… for schedule. First talk this Friday (i.e., tomorrow/today) by @florian_shkurti!
3
26
Prospective grad students: what factors do you consider when picking a grad school?
Here's a few things I wish I'd known when applying to gradschool
krrish94.github.io/blog/grad…
Thanks @adityakusupati @VikramVoleti for proofreading (2/n)
2
3
25
This is an *amazing* opportunity! Postdoc opening at @MontrealRobots
@duckietown_coo is a fantastic advisor; the group has access to excellent students at UdeM/Mila; diverse pool of collaborators too!
(Retweets greatly appreciated)
@UMontrealDIRO @Mila_Quebec @Montreal_AI
We have an opening for a postdoc in my group working on differentiable SLAM, sim2real, and uncertainty estimation for deep learning models deployed on robots. The form to apply is here: forms.gle/hiGxgDbUaCA929YZ6. Please share!
7
26
ConceptFusion produces 3D maps that are
(a) completely open-set
(b) inherently multimodal
These "queryable" scene representation can be used in downstream spatial perception tasks by simply tweaking the query and/or modality.
~5 min summary
piped.video/rkXgws8fiDs via @YouTube
1
5
26
7,079
Inviting submissions to our workshop @CVPR
multimodalitiesfor3dscenes.g…
See thread from @changanvr for more
I'm excited to announce the first Multimodalites for 3D Scenes (M3DS) workshop @CVPR : multimodalitiesfor3dscenes.g….
How do we understand 3D scenes and objects around us with multimodalities (vision👀, language🧾, audio👂, and touch🖐️)?
📢 Call for papers: submit before April 15!
5
25
8,981
In this new work led by Kei, we explore tactile sensing for determining extrinsic contact.
If you are at IROS, drop by the poster to chat with Kei!
Attending #IROS2023 at Detroit. I’m presenting our recent work “Tactile Estimation of Extrinsic Contact Patch for Stable Placement” at RoboTac2023 workshop.
Joint work with @devesh_dkj, @_krishna_murthy, @kanejaki, and Josh Tenenbaum.
arxiv.org/abs/2309.14552
4
25
7,094
Code for ConceptGraphs is now available at github.com/concept-graphs/co…
🚨 ConceptGraphs: Open-Vocabulary 3D Scene Graphs for Perception and Planning
RGB-D images -> 3D scene-graphs -> enhance with vision-language features
Enables a wide range of perception and planning abilities. Deployed in many real-world use-cases.
concept-graphs.github.io/
1
25
2,788
New work, led by Omama @padfoot7_ on autonomous driving without any HD maps.
Using openly available road network maps (openstreetmap) and a sparse set of approximate language landmarks, we develop an autonomous navigation system (1/N) 🧵
Check out our recent work:
ALT-Pilot: Autonomous Navigation with Language augmented Topometric maps
Takeaways:
- Crowdsourced augmentation of topometric maps with language landmarks
- Multimodal localization system for these enhanced maps
navigate-anywhere.github.io/…
2
2
26
7,259
From an admisisons committee perspective, the letters end up being not helpful. It's hard to compare two applicants *who've drafted their own letters*. (3/6)
1
1
23
More interesting work on multimodal queryable representations for robotics
Can robots 🤖 to navigate to sounds 🔊 they've heard?
w/ audio-language 🔊✏️ foundation models, excited that we can now ask our helper robots to "go to where you heard coughing"
Audio-Visual-Language Maps w/ @huang_chenguang @oier_mees @wolfram_burgard: avlmaps.github.io
22
2,378
Following @vincesitzmann's lead, we are handing out **two** ICCV registration tickets to students who might not otherwise have the financial support to attend.
DM me w/ a short blurb of why you'd like a ticket -- and we'll reach out with instructions!
Happening today!
"Differentiable 3D Vision and Graphics" workshop @ICCV_2021
Tune in to listen to an exciting speaker lineup - begins 1 pm EDT
Schedule: montrealrobotics.ca/diff3d/s…
ALT ICCV workshop on differentiable 3D vision and graphics features Yaron Lipman, Tzu-Mao Li, Nikhila Ravi, Wenzel Jakob, Katja Schwarz, Vincent Sitzmann, Clement Fuji-Tsang, Cengiz Oztireli, Marie-Julie Rakotosaona, and Matthew Tancik
1
4
24
Drafting an informative letter takes plenty of time, and experience. It's hard to expect a student (usually undergrad) to draft one! (2/6)
1
1
23
Registration for CS-CAN is now open!
Happening this June in Montreal @mcgillu ; multiple co-located conferences spanning robotics, vision (CRV), AI (CANAI), and other CS areas
cscan-infocan.ca/co-located-…
Exciting speaker lineup and events
1
7
21
7,299
Next speaker in the Robot Learning Seminar Series at
@MILAMontreal is Shubham Tulsiani @shubhtuls, on "Self-supervised interaction with the physical world"
Watch the livestream this Friday at 3 pm EDT here piped.video/watch?v=LWQ1QNT9…
Full schedule: montrealrobotics.ca/robotlea…
Announcing the robot learning seminar series at @MILAMontreal and @MontrealRobots. Great speaker lineup! See montrealrobotics.ca/robotlea… for schedule. First talk this Friday (i.e., tomorrow/today) by @florian_shkurti!
1
9
22
A neat thread -- esp for early PhD students 👇
Before grad school, I was really scared I wouldn't be "smart enough" to do well at research. So I want to share some traits & skills other than smarts that helped me. It's cool not to have all of these, everyone is different---just know it's not all about being "smart." 🧵
1
22
Looks like a very exciting framework to watch out for
We are excited to release Ivy, a new open source Deep Learning framework!
Ivy unifies the syntax and call signatures of existing frameworks.
Write your code once in Ivy, and support all frameworks simultaneously.
Links to Paper, Code and Docs here: ivy-dl.org
3
1
23
We're thrilled to have members of the ML, HCI, visualization and graphics communities. More speakers to join in soon
@lilianweng @hardmaru @FalaahArifKhan @hugo_larochelle @jeffbigham @evelynevs Terrence Parr
3/N We have an excellent line up of speakers & panelists including @lillian_weng @hardmaru @FalaahArifKhan @Hugo_larochelle @jeffbigham @evelynevs Terrence Parr and more to be confirmed soon.
1
22
It's surprising to see we still don't have a community-adopted solution to this in 2020. For example, I'd love to hover over the citation to see a tooltip listing the paper / authors
How about giving all conference submissions one extra page but insisting we use verbose citations, like (Mildenhall et al. 2020).
2
22
You might not realize, but dmissions committees know this and typically discount letters they suspect to be student-drafted (often they send follow-up requests to your referees). (4/6)
1
1
22
Advice to people writing a statement of purpose (a la @kkitani)
It's worth looking up the meaning of each word in the essay.
Read a statement a few weeks ago that went: "I painstakingly chose <insert-university-name> for my graduate study ..."
New blogpost up in the gradschool advice series
"Writing a statement of purpose"
If you're submitting grad school applications this week, might still come in handy :)
krrish94.github.io/blog/grad… (4/n)
1
2
22
The group of people @Mila_Quebec is just fantastic! And Canada in general has generous academic funding programs for all things AI.
If you're a robotics person on the faculty job market, a "must-apply"! :)
(Retweets appreciated)
@UMontrealDIRO @MontrealRobots @Montreal_AI
We have an open position in robotics at the assistant professor rank in the computer science department at University of Montreal (affiliated with Mila). Deadline to apply is Jan 31. Details here: diro.umontreal.ca/public/FAS…. Please help me and share widely!
5
21
I gave a talk last week, on how "classical" robotics techniques still fare better compared to learning methods. After the talk, I felt I should write my thoughts up, and here they are: medium.com/@krrish94/the-sta…
2
6
22
At RSS @rssconf these next few days. Reach out if you’re here too and would like to chat
21
1,665
New blogpost up in the gradschool advice series
"Writing a statement of purpose"
If you're submitting grad school applications this week, might still come in handy :)
krrish94.github.io/blog/grad… (4/n)
1
5
21
Excited to be presenting this at @icra2020. Thanks to fantastic collaborators: @mautkiungli @duckietown_coo
1
20