Miguel Jaques
Papers
1
Total Citations
3
H-Index
1
About
Miguel Jaques is a researcher at the forefront of computer vision and robotics, specializing in learning physical world representations from minimal supervision. His work bridges system identification, 3D perception, and dynamics modeling, with a focus on enabling machines to understand object motion and structure directly from raw video. In his landmark paper, "Vision-based system identification and 3D keypoint discovery using dynamics constraints" (2021), Jaques introduced V-SysId, a groundbreaking method that simultaneously discovers 3D keypoints, identifies system dynamics, and calibrates extrinsic camera parameters from a single unlabeled video—using only the object’s equations of motion as weak supervision. This work, which has garnered 3 citations, demonstrates a powerful paradigm for learning interpretable, physics-aware representations without costly manual annotations. Jaques’ contributions are particularly impactful for robotic manipulation, autonomous systems, and scene understanding, where extracting 3D structure and dynamics from limited data is critical. His approach exemplifies how combining geometric constraints with learning can unlock efficient, generalizable perception, making him a rising voice in the integration of vision and physics for intelligent systems.
Research Focus
Key Achievements
Top Papers
- 1