Mingyu You
Papers
11
Total Citations
95
H-Index
6
About
Mingyu You is a robotics and computer vision researcher whose work spans robot learning, 3D scene understanding, and autonomous navigation. His research is particularly focused on enabling robots to learn intelligently from human demonstrations and visual observations, reducing the costly reliance on real-world interactions that typically constrains robotic training pipelines. Among his most notable contributions is his work on 3D part assembly, where he developed transformer-based approaches to help robots understand and reconstruct complex object structures from individual components — a critical capability for autonomous manufacturing and household robotics. His paper on image-only imitation learning addresses a practical bottleneck in sim-to-real transfer, allowing robots to learn from expert demonstrations without expensive physical experimentation. You has also made meaningful advances in goal-conditioned reinforcement learning, introducing disentangled representations and reachability planning to help agents tackle long-horizon tasks more efficiently. Complementing these efforts, his research on dynamic dense CRF inference for video segmentation and semantic SLAM (21 citations) demonstrates a strong command of spatial perception fundamentals. With work spanning GAN-based movement primitives, contrastive video learning, and weakly supervised reinforcement learning, You's cumulative contributions reflect a coherent vision: building robots that learn robustly, efficiently, and naturally from human experience.
Research Focus
Key Achievements
Top Papers
- 1Dynamic dense CRF inference for video segmentation and semantic SLAM21 citations · 2022
- 23D Part Assembly Generation With Instance Encoded Transformer19 citations · 2022
- 3Robot learning from human demonstrations with inconsistent contexts13 citations · 2023
- 4
- 5
- 6GAN-Based Editable Movement Primitive From High-Variance Demonstrations7 citations · 2023
- 7
- 83D Assembly Completion5 citations · 2023
- 9Contrast, Imitate, Adapt: Learning Robotic Skills From Raw Human Videos3 citations · 2024
- 10Visual Landmark Learning Via Attention-Based Deep Neural Networks3 citations · 2021