PhD student @ Real Virtual Humans group (IMPRS-IS & @uni_tue) ex-Samsung AI
Tübingen, Germany
Joined April 2016
- Tweets16
- Following290
- Followers212
- Likes224
Pinned Tweet
🚀 Excited to share our #ICCV25 paper: TriDi (github.com/ptrvilya/tridi):
🎯 TL;DR: We generate 3D static Human, Object and Interaction simultaneously through a three-way diffusion process, modeling seven distributions with one network.
Thanks to @yxue_yxue and Margaret for providing the registrations for our visualizations! Checkout their work on high-fidelity human registration!
Amazing results, very entertaining to watch the teaser ;)
We proudly power the human appearance: yuxuan-xue.com/avaimg/
A Blender Add-On for unifying 15 HOI datasets: yuxuan-xue.com/hoi-blender/
More frontier exciting research coming soon!
Glad to share COSMI: a scalable way to compose single-object interactions into multi-object sequences. I really enjoyed collaborating on this with @daniel_eskandar. He is looking for PhD positions and would be an excellent candidate!
🌐 ptrvilya.github.io/cosmi/
Capturing every combination of human-object interaction is not feasible. We need compositionally.
Most interactions are local. A hand holds a cup, a chair supports the pelvis, while much of the body remains free.
COSMI, by Daniel Escandar @ptrvilya builds on this observation to compose existing single-object datasets into 222k multi-object sequences, totaling 275 hours, with up to 5 objects per sequence. LLM reasoning and geometric checks ensure the compositions remain plausible.
On this data, we train a single diffusion transformer for interactions with 1 to 5 objects, predicting each object relative to its interacting body part and generalizing to unseen objects and combinations.
Data and code will be made available:
🌐 lnkd.in/eHdHrFbt
🎉 Excited to be at #ECCV2026 🇸🇪 . I'll be presenting two posters here, so come say hi if you are around!
📄 ECHO: Ego-centric Modeling of Human-Object Interactions
📍 Friday 11.09 10:30-12:00 - Poster #267
🤝Joint work with @guzov_vladimir @_R_Marin_ Emre Aksan, Xu Chen, Daniel Cremers, Thabo Beeler @GerardPonsMoll1
📄 TrustCLIP: Learning Private Visual Features via Adversarial Reconstruction
📍 Saturday 12.09 10:30-12:00 - Poster #386
🤝Joint work with @_nikos_athan Angela Yao, Shugao Ma, Eric Sauser Edoardo Remelli, Shreyas Hampali, Johannes Schönberger, Fadime Sener, Bugra Tekin
🖼️ Stop by our poster (Poster Session 2) in Honolulu 🌴, where @anticdimi has kindly agreed to cover my absence :)
Joint work with @_R_Marin_ @ChibaneJulian @GerardPonsMoll1
Blendify 2.0 is out!🎉
This major release of our Python framework for creating and rendering scenes in Blender introduces several new features. More details below 🧵 (1/3).
GitHub: github.com/ptrvilya/blendify
Developed together with @guzov_vladimir.
(2/3) Blendify 2.0 has:
- a flexible way to define materials with more material presets;
- support for common ways to define rotation;
- native support for look_at rotation mode to ease camera positioning;
- improved handling of lights;
- fast preview rendering using eevee.
(3/3) For a deeper dive, check out the feature walkthrough: virtualhumans.mpi-inf.mpg.de…
and read our technical report: arxiv.org/abs/2410.17858
Looking forward to your participation in our #CVPR24 @CVPR workshop on Reconstruction of Human-Object Interactions. This year we return with 5 challenges and an amazing speaker lineup!
More info at: rhobin-challenge.github.io/
Excited to help build this amazing course!
Our virtual humans lecture videos on YouTube:
youtube.com/playlist?list=PL…
We cover from the basics (image formation, surface reps), to classical models like SMPL to modern NeRF based ones, as well as models of human motion in scenes based on transformers, diffusion, etc. Enjoy!
So happy to finally release our framework🥳
Me and @ptrvilya are excited to release Blendify 🎉 – a Python framework for creating and rendering scenes in Blender. It is useful for various computer vision visualizations – No need to dive into Blender API!
Link: github.com/ptrvilya/blendify
More details below 🧵 (1/4):
Ilya Petrov retweeted
Call for submissions: 1st workshop on reconstruction of human-object interactions. #CVPR2023 @CVPR . We invite submissions related to modelling human object interactions in 3D, 2.5D, 2D, images, videos and more! Deadline: March 20.
More details: rhobin-challenge.github.io/#…
Ilya Petrov retweeted
I am proud and delighted to be able to work with researchers who take reviewing seriously! 5 people in my group recognized for their efforts!
Congratulations @ChibaneJulian @_garvita_tiwari @bharat_b7 @ptrvilya @_R_Marin_ !
From our group, Bharat Lal Bhatnagar, Garvita Tiwari, Ilya Petrov, Riccardo Marin, and I are Outstanding Reviewers at CVPR22! cvpr2022.thecvf.com/outstand…
Also, many other colleagues are acknowledged! Congrats everyone!
Ilya Petrov retweeted
BEHAVE: The first dataset with reliable 3D Human + Objects paired with images / depth.
If you work on 3D human pose and shape from images/ video/depth/multi-view, interaction modelling, human-scene interaction, motion forecasting, check out BEHAVE!
virtualhumans.mpi-inf.mpg.de…