|
I am a Ph.D. student in the GRASP laboratory at the University of Pennsylvania advised by Kostas Daniilidis. I am interested in deep learning, computer vision, and robotics. Most of my recent work concerns deep predictive models of videos. I received my bachelor's degree from Czech Technical University in Prague, where I worked with Tomas Pajdla. I've spent time at INRIA with Josef Sivic, TiTech with Akihiko Torii, and UC Berkeley with Sergey Levine and Chelsea Finn. My name is pronounced as "Oleg". I also prefer being called that in less formal writing. Wikipedia tries to explain the pronunciation here. Google Scholar / GitHub / Email / CV / LinkedIn |
|
|
|
I am interested in building agents that are capable of predicting the future, and using this prediction capability to act in the world. I believe that using vision as sensing modality is crucial for making such agents general purpose, and testing these algorithms on a real robotic system is one of the only sure ways to make progress towards intelligence. My recent work in this area involves machines trying to understand agent motion, physics, interesting moments in time, and human behavior, as well as intrinsically motivated machines. During my bachelor's, I worked on camera geometry for structure from motion and proposed an algorithm for robust estimation of camera focal length. Check out this and my other fun projects on my GitHub page. |
 
|
We propose a hierarchical predictive model that predicts a sequence starting from the high level events and progressively fills in finer and finer details. We train the model on goal-conditioned prediction on up to 80-frames (=12.5 seconds) videos.
|
 
|
We learn to infer an action representation from either motor or sensory input by using a dual variational autoencoder. By learning a dynamics model in such semi-supervised manner, we achieve both high data efficiency and planning performance.
|
 
|
We learn a latent variable model for dynamics of image observations, and use it to construct an agent that maximizes Bayesian surprise of the future frames. The Bayesian agent can perform exploration that is more robust in stochastic environments than simpler prior prediction schemes.
|
 
|
We discover keyframes in videos by learning to select frames that enable prediction of the entire sequence. We show that our method improves performance of hierarchical planning by finding meaningful keyframes in demonstration data. Hover the mouse (or tap the screen) here to see the video. |
 
|
We learn to discover an agent's action space along with a dynamics model from pure video data. After a calibration stage, the model can be used to perform model predictive control, requiring orders of magnitude fewer action-annotated videos than other methods. Hover the mouse (or tap the screen) here to see the video. |
 
|
The model predicts future video frames by learning to represent the present state of a system together with a high-level transformation that is used to produce its future state. Hover the mouse (or tap the screen) here to see the video. |
|
|
|
 
|
MSE loss and its variants are commonly used for training and evaluation of future prediction. But is this the right thing to do? Hover the mouse (or tap the screen) here to see the video. |
|
Books:
|
|
I am actively looking for students who are strongly motivated to work on a research project, including students who want to do a Master's thesis. Check out some of my work above and if you find it interesting, do send me an email!
|
|
|