Mara Levy

I am a PhD student in the Computer Science department at the University of Maryland, College Park. I work on robotic learning and computer vision with Abhinav Shrivastava. I graduated Summa Cum Laude from the University of Pennsylvania where I majored in Computer Science and minored in Mathematics.

Email  /  CV  /  Scholar  /  Github

Research

My research utilizes a mixture of computer vision and robotics approaches to try and make robots more useful in the real world. Currently I am interested in state representation for robotic learning. In the long term I am interested in using these techniques to learn from online videos of humans.

Using keypoints to train generalized robot policies on a wide variety of robotics tasks.

A novel framework that integrates few-shot expert demonstrations with a tri-teaching strategy for effective noise mitigation in preference based reinforcement learning.

This approach generates a scene graph from input images and uses it to iteratively check and correct action sequences generated by an LLM-based task planner.

A novel method that is capable of teaching a policy to interact with objects that are not present in the dataset through the use of NERF based scene augmentation.

A novel method for learning complex goal-conditioned robotics tasks from a single demonstration, using unique reward function and knowledge expansion.

A novel method for learning complex goal-conditioned robotics tasks from a single demonstration, using unique reward function and knowledge expansion.

A method for projecting poses into an embedding space that is view invariant. This means the same pose shot from different camera angles should have an equal embedding.

Build efficient Transformer design for non-parametric human mesh recovery with coarse-to-fine pipeline.

A planning algorithm that helps reinforcement learning algorithms that can typically only solve tasks for static objects solve the same task, but with a dynamic object.

Template Credits Jon Barron