Research Territories
Our research focuses on document understanding, autonomous agents, and visual systems for robotics.
Document Understanding
Our document understanding research focuses on improving end-to-end models for parsing forms and documents. We leverage synthetic ground truth pairs of commonly filled form data to train more robust models. Published at WACV VisionDocs.
Explore research →Computer Use
We're building and expanding the foundational components critical to enabling long-horizon autonomous computer use tasks. This includes agent infrastructure, task planning, and robust execution systems.
Explore research →Visual Odometry
Our aerospace research focuses on data collection and sensor fusion for visual inertial odometry on dynamical kinetic systems (drones). This enables robust state estimation for autonomous flight.
Explore research →Synthetic Data
We develop techniques for generating high-quality synthetic training data, particularly focused on document and form understanding. Our methods improve model robustness and reduce reliance on expensive manual annotation.
Explore research →Propose New Research
Have an idea for a new research direction? We welcome proposals from the community. All research directions are discussed and decided collectively by contributors.
Learn How to Contribute