Papers
10
Total Citations
471
H-Index
9
About
Judy Hoffman is a prominent researcher at the intersection of computer vision, robotics, and machine learning, with a particular focus on domain adaptation, embodied AI, and robot learning. Her work addresses one of the field's most persistent challenges: enabling models trained in one environment to perform reliably in another. Her most cited contribution, "VisDA: A Synthetic-to-Real Benchmark for Visual Domain Adaptation" (2018, 181 citations), established a foundational benchmark for evaluating how well models transfer from synthetic to real-world imagery — a critical problem when real labeled data is costly to obtain. This theme extends into her robotics work, where papers like "Towards Adapting Deep Visuomotor Representations from Simulated to Real Environments" (61 citations) explore how robots can leverage simulation for training and adapt effectively to physical deployment. More recently, Hoffman has pushed into open-world navigation with "ZSON: Zero-Shot Object-Goal Navigation" (41 citations) and scalable robot imitation learning through her "EgoMimic" framework, which leverages egocentric human video for manipulation tasks. Spanning motion planning, object detection, and embodied intelligence, her body of work reflects a sustained commitment to making AI systems more adaptable, practical, and deployable in real-world settings.
Research Focus
Key Achievements
Top Papers
- 1VisDA: A Synthetic-to-Real Benchmark for Visual Domain Adaptation181 citations · 2018
- 2Adapting Deep Visuomotor Representations with Weak Pairwise Constraints81 citations · 2020
- 3
- 4ZSON: Zero-Shot Object-Goal Navigation using Multimodal Goal Embeddings41 citations · 2022
- 5
- 6
- 7
- 8Interactive adaptation of real-time object detectors15 citations · 2014
- 9EgoMimic: Scaling Imitation Learning via Egocentric Video9 citations · 2025
- 10EgoMimic: Scaling Imitation Learning via Egocentric Video3 citations · 2024