Ilya A. Petrov

I'm a final-year PhD candidate at the Tübingen AI Center, supervised by Gerard Pons-Moll. My research is in 3D computer vision and digital humans, focusing on modeling human-object interactions, human motion, and egocentric vision.

I obtained my B.Sc. and M.Sc. from Moscow State University, where I worked on human shape and pose reconstruction from images. Before starting my PhD, I worked at Samsung AI Center on interactive image segmentation.

Selected Publications

ECHO: Ego-Centric modeling of Human-Object interactions

Ilya A. Petrov, Vladimir Guzov, Riccardo Marin, Emre Aksan, Xu Chen, Daniel Cremers, Thabo Beeler, Gerard Pons-Moll

ECCV 2026

Website arXiv Code

Reconstructing full-body human-object interactions from sparse egocentric tracking of the head and wrists.
TrustCLIP: Learning Private Visual Features via Adversarial Reconstruction

Nikos Athanasiou, Ilya A. Petrov, Angela Yao, Shugao Ma, Eric Sauser, Edoardo Remelli, Shreyas Hampali, Johannes Schönberger, Fadime Sener, Bugra Tekin

ECCV 2026

Website arXiv

Training a projection of CLIP-style features against a generative reconstruction attacker, keeping them useful for downstream tasks but no longer invertible into recognizable images.
TriDi: Trilateral Diffusion of 3D Humans, Objects, and Interactions

Ilya A. Petrov, Riccardo Marin, Julian Chibane, Gerard Pons-Moll

ICCV 2025

Website Video arXiv Code

A single diffusion model of the joint distribution of 3D humans, objects, and their interactions, conditioned on any subset of the three.
Object pop-up: Can we infer 3D objects and their poses from human interactions alone?

Ilya A. Petrov, Riccardo Marin, Julian Chibane, Gerard Pons-Moll

CVPR 2023

Website Video arXiv Code

Inferring the 3D object a person is interacting with from the human body alone.
BEHAVE: Dataset and Method for Tracking Human Object Interactions

Bharat Lal Bhatnagar, Xianghui Xie, Ilya A. Petrov, Cristian Sminchisescu, Christian Theobalt, Gerard Pons-Moll

CVPR 2022

Website Video arXiv Data

A multi-view RGB-D dataset of humans interacting with everyday objects, with a method for jointly tracking both.
F-BRS: Rethinking Backpropagating Refinement for Interactive Segmentation

Konstantin Sofiiuk, Ilya A. Petrov, Olga Barinova, Anton Konushin

CVPR 2020

Video arXiv

Fast click-based interactive segmentation that refines predictions by optimizing a small set of auxiliary parameters instead of the network input.

Other Projects

Blendify

Python rendering framework for Blender.

Code Website Report