Seminar: Graduate Seminar
Advances in 3D Vision: Image-Based Scene and Mesh Understanding
Lecturer:
Amir-Eli Belder
Research Areas:
| This thesis explores advances in 3D vision through image-based scene understanding and 3D mesh analysis, using modern approaches such as multimodal attention, mixture-of-experts architectures, and reinforcement learning for 3D perception and autonomous decision-making in complex environments. The work focuses on enabling intelligent systems to better perceive, navigate, and understand visual and geometric information.
In the domain of scene understanding, the research investigates efficient navigation and destination search from egocentric visual inputs, aiming to improve spatial understanding and assistive navigation systems. In parallel, the thesis studies geometric understanding of 3D meshes, including adversarial robustness in 3D deep learning models and advanced feature fusion methods for enhanced shape understanding. Overall, this work contributes toward robust and intelligent 3D vision systems for applications in robotics, autonomous agents, and spatial AI. |
| Ph.D. student under the supervision of Prof. Ayelet Tal.
|