Katerina Fragkiadaki

I am interested in building machines that understand the stories that videos portray, and, inversely, in using videos to teach machines about the world. The pen-ultimate goal is to build a machine that understands movie plots, and the ultimate goal is to build a machine that would want to watch Bergman over this.

I am a JPMorgan Chase Associate Professor of Computer Science in the Machine Learning Department at Carnegie Mellon University. Prior to joining MLD's faculty I spent three wonderful years as a post doctoral researcher first at UC Berkeley working with Jitendra Malik and then at Google Research in Mountain View working with the video group. I completed my Ph.D. in GRASP, UPenn with Jianbo Shi. I did my undergraduate studies at the National Technical University of Athens and before that I was in Crete.

Dear prospective graduate students: There is no need to email, please submit your application to our School following the links here: MLD Ph.D. or R.I. Ph.D.

8107 Gates Hillman
Carnegie Mellon University
Pittsburgh, PA, 15213
katef 'at' cs.cmu.edu
Admin: Christine Martik, cmartik@andrew.cmu.edu



Deep Reinforcement Learning and Control Spring 2024
Deep Reinforcement Learning and Control Fall 2023
Deep Reinforcement Learning and Control Spring 2023
Deep Reinforcement Learning and Control Fall 2022


Teaching machines to appreciate Bergman is a hard problem, in fact, most people fail to do so. Thus, it is a very long term goal of our research. Meanwhile, we are working on teaching machines basic common sense. Our recent works use recurrent nets with 3D representation bottlenecks that disentangle camera motion from appearance, while being optimized end-to-end for a final task, such as view prediction or 3D object detection. In this way, our models learn object permanence, size constancy, occlusions and dis-occlusion relationships, useful for 3D object detection, affordance learning and language grounding.

Selected Publications

Video Diffusion Alignment via Reward Gradients
Mihir Prabhudesai, Russell Mendonca, Zheyang Qin, Katerina Fragkiadaki, Deepak Pathak

VADER alignes video diffusion models using end-to-end reward gradient backpropagation from off-the-shelf differentiable reward functions.
paper | project page
ICAL: Continual Learning of Multimodal Agents by Transforming Trajectories into Actionable Insights
Gabriel Sarch, Lawrence Jang, Michael Tarr, William Cohen, Kenneth Marino, Katerina Fragkiadaki

ICAL is a SGD-free in-context VLM method where VLMs are promopted to map demonstration trajectories into experience abstractions to store in an external memory, and retrieve them on-the-fly to use as in-context examples.
paper | project page
DreamScene4D: Dynamic Multi-Object Scene Generation from Monocular Videos
Wen-Hsuan Chu*, Lei Ke*, Katerina Fragkiadaki

DreamScene4D generates 3D dynamic scenes of multiple objects from monocular videos training-free, using object-centric diffusion priors and pixel and motion reprojection error.
paper | project page
ODIN: A Single Model for 2D and 3D Perception
Ayush Jain, Pushkal Katara, Nikolaos Gkanatsios, Adam W. Harley, Gabriel Sarch, Kriti Aggarwal, Vishrav Chaudhary, Katerina Fragkiadaki

ODIN processes both RGB images and sequences of posed RGB-D images by alternating between 2D and 3D fusion layers using projection and unprojection from camera info. New SOTA in Scannet200.
CVPR 2024
paper | project page with code
3D Diffuser Actor: Policy Diffusion with 3D Scene Representations
Tsung-Wei Ke, Nikolaos Gkanatsios, Katerina Fragkiadaki

Combining 3D relative attention transformers with action trajectory diffusion gives SOTA imitation learning robot policies in CALVIN and RLbench.
paper | project page with code
Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous Driving and Zero-Shot Instruction Following
Brian Yang, Huangyuan Su, Nikolaos Gkanatsios, Tsung-Wei Ke, Ayush Jain, Jeff Schneider, Katerina Fragkiadaki

We combine trajectory diffusion models with evolutionary search and achieve SOTA performance in nuPLAN. We prompt LLMs to map language instructions to shaped reward functions, and optimize them with diffusion-ES, and solve the hardest driving scenarios.
CVPR 2024
paper | project page with code
Test-time Adaptation of Discriminative Models via Diffusion Generative Feedback
Mihir Prabhudesai, Tsung-Wei Ke, Alexander C. Li, Deepak Pathak, Katerina Fragkiadaki
NeurIPS 2023
paper | project page with code
Open-Ended Instructable Embodied Agents with Memory-Augmented Large Language Models
Gabriel Sarch, Yue Wu, Michael J. Tarr, Katerina Fragkiadaki
EMNLP findings 2023
paper | project page with code
Act3D: 3D Feature Field Transformers for Multi-Task Robotic Manipulation
Theophile Gervet, Zhou Xian, Nikolaos Gkanatsios, Katerina Fragkiadaki
CoRL 2023
paper | project page with code
Gen2Sim: Scaling up Robot Learning in Simulation with Generative Models
Pushkal Katara, Zhou Xian, Katerina Fragkiadaki
ICRA 2024
paper | project page with code
ChainedDiffuser: Unifying Trajectory Diffusion and Keypose Prediction for Robotic Manipulation
Zhou Xian, Nikolaos Gkanatsios, Theophile Gervet, Tsung-Wei Ke, Katerina Fragkiadaki
CoRL 2023
paper | project page with code
Test-time Adaptation with Slot-Centric Models
Mihir Prabhudesai, Anirudh Goyal, Sujoy Paul, Sjoerd van Steenkiste, Mehdi S. M. Sajjadi, Gaurav Aggarwal, Thomas Kipf, Deepak Pathak, Katerina Fragkiadaki
ICML 2023
paper | project page with code
Energy-based Models are Zero-Shot Planners for Compositional Scene Rearrangement
Nikolaos Gkanatsios, Ayush Jain, Zhou Xian, Yunchu Zhang, Christopher Atkeson, Katerina Fragkiadaki
RSS 2023
paper | project page with code
Simple-BEV: What Really Matters for Multi-Sensor BEV Perception?
Adam W. Harley, Zhaoyuan Fang, Jie Li, Rares Ambrus, Katerina Fragkiadaki
ICRA 2023
paper | project page with code
FluidLab: A Differentiable Environment for Benchmarking Complex Fluid Manipulation
Zhou Xian, Bo Zhu, Zhenjia Xu, Hsiao-Yu Tung, Antonio Torralba, Katerina Fragkiadaki, Chuang Gan
ICLR 2023, spotlight
paper | project page with code
Analogy-Forming Transformers for Few-Shot 3D Parsing
Nikolaos Gkanatsios, Mayank Singh, Zhaoyuan Fang, Shubham Tulsiani, Katerina Fragkiadaki
ICLR 2023
paper | project page with code
Bottom Up Top Down Detection Transformers for Language Grounding in Images and Point Clouds
Ayush Jain, Nikolaos Gkanatsios, Ishita Mediratta, Katerina Fragkiadaki
ECCV 2022
paper | project page with code
TIDEE: Tidying Up Novel Rooms using Visuo-Semantic Commonsense Priors
Gabriel Sarch, Zhaoyuan Fang, Adam W. Harley, Paul Schydlo, Michael J. Tarr, Saurabh Gupta, Katerina Fragkiadaki
ECCV 2022
paper | project page with code
Particle Videos Revisited: Tracking Through Occlusions Using Point Trajectories
Adam W. Harley, Zhaoyuan Fang, Katerina Fragkiadaki
ECCV 2022, oral
paper | project page with code
Visually-Grounded Library of Behaviors for Manipulating Diverse Objects across Diverse Configurations and Views
Jingyun Yang*, Hsiao-Yu Fish Tung*, Yunchu Zhang*, Gaurav Pathak, Ashwini Pokle, Christopher G Atkeson, Katerina Fragkiadaki
CoRL 2021
paper | project page with code
Disentangling 3D Prototypical Networks for Few-Shot Concept Learning
Mihir Prabhudesai*, Shamit Lal*, Darshan Patil*, Hsiao-Yu Tung, Adam Harley, Katerina Fragkiadaki
ICLR 2021
paper | project page with code
HyperDynamics: Generating Expert Dynamics Models by Observation
Zhou Xian, Shamit Lal, Hsiao-Yu Tung, Emmanouil Antonios Platanios, Katerina Fragkiadaki
ICLR 2021
paper | project page with code
Track, Check, Repeat: An EM Approach to Unsupervised Tracking
Adam W. Harley, Yiming Zuo, Jing Wen, Ayush Mangal, Shubhankar Potdar, Ritwick Chaudhry, Katerina Fragkiadaki
CVPR 2021
paper | project page with code
Move to See Better: Self-Improving Embodied Object Detection
Zhaoyuan Fang, Ayush Jain, Gabriel Sarch, Adam W. Harley, Katerina Fragkiadaki
BMVC 2021
paper | project page
CoCoNets: Continuous Contrastive 3D Scene Representations
Shamit Lal, Mihir Prabhudesai, Ishita Mediratta, Adam W. Harley, Katerina Fragkiadaki
CVPR 2021
paper | project page with code
Tracking Emerges by Looking Around Static Scenes, with Neural 3D Mapping
Adam W. Harley, Shrinidhi K. Lakshmikanth, Paul Schydlo, Katerina Fragkiadaki
ECCV 2020
Embodied Language Grounding with Implicit 3D Visual Feature Representations
Mihir Prabhudesai*, Hsiao-Yu Fish Tung*, Syed Ashar Javed*, Maximilian Sieb, Adam W. Harley, Katerina Fragkiadaki
CVPR 2020
paper | project page
Epipolar Transformers
Yihui He, Rui Yan, Katerina Fragkiadaki, Shoou-I Yu
CVPR 2020
paper | project page
Graph-structured Visual Imitation
Xian Zhou*, Max Sieb*, Audrey Huang, Oliver Kroemer, Katerina Fragkiadaki
CoRL 2019
paper | project page with code
Learning from Unlabelled Videos Using Contrastive Predictive Neural 3D Mapping
Adam W. Harley, Fangyu Li, Shrinidhi K. Lakshmikanth, Xian Zhou, Hsiao-Yu Fish Tung, Katerina Fragkiadaki
ICLR 2020
paper | project page with code
Learning Spatial Common Sense with Geometry-Aware Recurrent Networks
Hsiao-Yu Fish Tung, Ricson Cheng, Katerina Fragkiadaki
CVPR 2019, oral presentation
paper | project page with code
Model Learning for Look-ahead Exploration in Continuous Control
Arpit Agarwal, Katharina Muelling and Katerina Fragkiadaki
AAAI 2019, oral presentation
paper | project page with code
Data Dreaming for Object Detection: Learning Object-Centric State Representations for Visual Imitation
Maximilian Sieb and Katerina Fragkiadaki
Humanoids 2018, oral presentation
paper | slides
Reinforcement Learning of Active Vision for Manipulating Objects under Occlusionss
Ricson Cheng, Arpit Agarwal, and Katerina Fragkiadaki
CoRL 2018
paper | slides | code
Geometry-Aware Recurrent Neural Networks for Active Visual Recognition
Ricson Cheng, Ziyan Wang, and Katerina Fragkiadaki
NIPS 2018
Reward Learning from Narrated Demonstrations
Fish Tung, Adam Harley, Liang-Kang Huang, Katerina Fragkiadaki
CVPR 2018
paper | bibtex
Depth-adaptive Computational Policies for Efficient Visual Tracking
Chris Ying, Katerina Fragkiadaki
paper | bibtex
Self-supervised Learning of Motion Capture
Hsiao-Yu Fish Tung, Wei Tung, Ersin Yumer, Katerina Fragkiadaki
NIPS 2017, spotlight
paper | bibtex | code
Adversarial Inverse Graphics Networks: Learning 2D-to-3D Lifting and mage-to-Image Translation from Unpaired Supervision
Hsiao-Yu Fish Tung, Adam Harley, William Seto and Katerina Fragkiadaki
ICCV 2017
paper | bibtex | code
SfM-Net: Learning of Structure and Motion from Video
Sudheendra Vijayanarasimhan, Susanna Ricco, Cordelia Schmid, Rahul Sukthankar and Katerina Fragkiadaki
Motion Prediction Under Multimodality with Conditional Stochastic Networks
Katerina Fragkiadaki, Jonathan Huang, Alex Alemi, Sudheendra Vijayanarasimhan, Susanna Ricco and Rahul Sukthankar
arxiv | video results
Learning Feature Hierarchies from Long-Range Temporal Associations in Videos
Panna Felsen, Katerina Fragkiadaki, Jitendra Malik and Alexei Efros
Workshop on Transfer and Multi-task Learning, in conjunction with NIPS 2015
Learning Predictive Visual Models of Physics for Playing Billiards
Katerina Fragkiadaki*, Pulkit Agrawal*, Sergey Levine and Jitendra Malik
ICLR 2016
paper | project page
Recurrent Network Models for Human Dynamics
Katerina Fragkiadaki, Sergey Levine, Panna Felsen and Jitendra Malik
ICCV 2015
paper | project page
Human Pose Estimation with Iterative Error Feedback
Joao Carreira, Pulkit Agrawal, Katerina Fragkiadaki, and Jitendra Malik
paper|project page with code
Learning to Segment Moving Objects in Videos
Katerina Fragkiadaki, Pablo Arbelaez, Panna Felsen and Jitendra Malik
CVPR 2015
paper | poster | bibtex | project page with code
Grouping-Based Low-Rank Video Completion and 3D Reconstruction
Katerina Fragkiadaki, Marta Salas, Pablo Arbelaez, and Jitendra Malik
NIPS 2014
paper | poster | bibtex | project page with code
Two Granularity Tracking: Mediating Trajectory and Detection Graphs for Tracking under Occlusions
Katerina Fragkiadaki, Weiyu Zhang, Geng Zhang, and Jianbo Shi
ECCV 2012
paper | poster | bibtex | project page with code
Video Segmentation by tracing Discontinuities in a Trajectory Embedding
Katerina Fragkiadaki, Geng Zhang, and Jianbo Shi
CVPR 2012
paper | poster | bibtex | project page with code
Detection-free Tracking: Exploiting Motion and Topology for Segmenting and Tracking under Entanglement
Katerina Fragkiadaki and Jianbo Shi
CVPR 2011
paper | poster | bibtex

Current Students

Wen-hsuan Chu R.I. Ph.D.
Mihir Prabhudesai R.I. Ph.D., (co-advised with Deepak Pathak)
Xian Zhou R.I. Ph.D.
Gabriel Sarch , Neuroscience Institute and Center for the Neural Basis of Cognition Ph.D. student, co-advised with Mike Tarr
Brian Yang R.I. Ph.D., co-advised with Jeff Schneider
Nikos Gkanatsios R.I. Ph.D.
Ayush Jain R.I. MSR
Theo Gevret MLD Ph.D.
Tsung-Wei Ke PostDoc


Fish Tung, MLD Ph.D., now post doc in M.I.T.
Adam Harley, R.I. Ph.D.
Ricson Chen, undergrad CSD
Zhaoyuan Fang, MSR, R.I.
Mayank Singh, MSR, R.I.
Yunchu Zhang, MSR, R.I.
Ziyan Wang, MSR
Shamit Lal, MSCV
Yiming Zuo , MSR, now Ph.D. in Princeton
Max Sieb, MSR
Arpit Agarwal, MSR
Henry Huang, MSML
Chris Ying, MSML
Yijie Wang, MSML
Darshan Patil , undergraduate CSD, now in MILA Ph.D.
Gaurav Pathak, research associate
Shrinidhi Kowshika Lakshmikanth, research associate
Ashwini Polke, MLD Ph.D.