Omid Taheri
@omidtaherii
Followers
517
Following
1K
Media
14
Statuses
89
Postdoc Researcher @PerceivingSys, MPI-IS. Cooking realistic virtual humans.
Germany
Joined January 2020
Another great work from our lab on reconstructing human-object-interaction from a single image. Don't miss it. #CVPR2025.
If you’re not in freefall, you’re likely contacting something. Yet 3D human-object interaction (HOI) reconstruction remains underexplored. PICO (#CVPR2025) recovers humans 🏃♂️, objects 🏓, and their interactions 👉🍎- all in 3D, from just a single internet image. 1/11
0
1
14
Proud of this work, make sure to check it out! #CVPR2025.We use VLMs to accurately infer 3D contacts on both humans and objects, the first method that generalizes to "in-the-wild" images, and achieves SOTA results.
InteractVLM (#CVPR2025) is a great collaboration between @MPI_IS, @UvA_Amsterdam and @Inria. Authors: @saidwivedi, @anticdimi, @sha2nk_t, @omidtaherii, @CordeliaSchmid, @Michael_J_Black and @dimtzionas.Code & models available at: (10/10)
0
0
13
RT @dimtzionas: 🎯 Modeling how people interact with objects in #3D is important for both 3D #perception from 2D images, and for 3D #avatar….
0
16
0
RT @DanecekRadek: 1/n Proud to present our latest attempt to give 3D faces a voice: “THUNDER: Supervising 3D Talking Head Avatars with Anal….
0
12
0
RT @dimtzionas: Reconstructing 3D human-object interactions from images/videos is a key step towards realistic #3D humans for #AR/#VR and a….
0
21
0
Looking forward to #ECCV in Milan! Happy to chat about research, collaboration, career, or anything else. We're also searching for talented interns-R2 rated the level of skills we're looking for proportioned to the cheese 🍕 in the video! How much cheese do you get? Let's talk 😁
0
0
7
What a day! Just defended my PhD with summa cum laude🎉 Huge thanks to my amazing advisors, @Michael_J_Black @dimtzionas, and committee @GerardPonsMoll1 @angelaqdai @AutoVisionGroup. Best feeling ever, esp. when they said it was the best presentation they’ve seen! #PhDone.@MPI_IS
Congratulations to @omidtaherii on defending his PhD summa cum laude on the topic of “Modeling Dynamic 3D Human-Object Interactions: From Capture to Synthesis”. Omid’s work on the GRAB dataset, GRIP, GOAL, and more has changed the field of human-object interaction.
17
5
94
A week ago had a discussion with a colleague who argued that AI doesn't need physics and can reason from pixels alone. This is a great example showing AI needs to learn physics and 3D, not just pixels. The real world works in 3D, and AI should reflect that for realistic results.
CW: Body Horror? . This AI video attempt to show gymnastics is one of the best examples I have seen that AI doesn’t actually understand the human body and it’s motion but is just regurgitating available data. (Which appears to be minimal for gymnastics)
0
0
6
AGI might still be a long way off :D, unless you can't understand this image in under a minute.
2
0
2
*Authors*: To the best of our knowledge, no other method generates such physically plausible motions. *Generated Motions*:
7
2
85
RT @Marilyn59846278: Wondering what people look like below the skin and fat? 💪🩻💀We introduce HIT: Human Implicit Tissues .
0
34
0
RT @saidwivedi: 🚀 With our TokenHMR work at #CVPR2024 🎉, you no longer need to compromise 3D accuracy at the expense of 2D image alignment….
0
10
0
RT @Michael_J_Black: WANDR (#CVPR2024) automatically generates natural 3D human motions given goals like reaching for an object. WANDR lear….
0
16
0
Preparing my PhD defense slides these days and thrilled about AI's progress! When I started, hardly anyone studied human-object interactions—mostly just body motion. I doubted myself sometimes, but now the impact is clear! 🚀 Don't be afraid to work on what others aren't! 💡.
It's going 3D! But with real humans, interactions and behaviors ;).
2
2
18
No more datasets without finger motions 🚫 With GRIP, getting accurate object-interaction datasets, including finger motions, is made easier ✨. Just capture body and object motions, and let GRIP fill in realistic finger motions. Find out more at #3DV2024.
To learn generative models of human-object interaction we need motion capture with fingers. Fingers are hard to capture so most datasets only contain the body and object motion. GRIP takes such body and object motions (without fingers) and adds realistic finger motions. (1/6)
0
6
27
RT @partha_ghosh0: 🚀 Ever wished your photo could come to life? Now it can! Transform your image into a dynamic avatar that speaks with yo….
0
8
0
RT @sha2nk_t: 📢📢 The AMASS version of our MOYO dataset is now online! MOYO contains complex yoga motions performed by a professional yoga i….
0
21
0
RT @zc_alexfan: Feel free to drop by our HANDS👋 workshop at #ICCV2023 13:30-17:30 Monday, room W5 for our amazing speakers @HughWang19, @gu….
0
12
0
4 long years of missing out on conferences during my PhD (thanks-visa issues), finally no visa needed for #ICCV23 (Europe perks). Feeling like I’ve got some networking catch-up to do😅. DM me if you are up for connecting and sharing ideas. Excited to meet many of you. #Networking
2
2
37
Don't miss this golden opportunity with Dimitris, an exceptional advisor. His mentorship transformed my research, and his insights are consistently impressive. Highly recommended! #PhDPosition.
0
0
2