Vincent Casser

Harvard University Press, Nomor Research (Germany)

Papers

Total Citations

493

H-Index

About

Vincent Casser is a leading researcher at the intersection of computer vision, robotics, and autonomous systems, with a primary focus on 3D scene understanding and multi-modal perception. His most influential work, "Depth Prediction without the Sensors," has garnered 491 citations and pioneered unsupervised learning of scene depth and ego-motion from monocular videos—a breakthrough that eliminates the need for expensive depth sensors, making robot navigation more accessible and cost-effective. This work leverages structural cues in video sequences to predict depth directly from RGB inputs, a critical capability for both indoor and outdoor navigation. More recently, Casser has advanced the field of instance segmentation with his cross-modal consistency approach (2022), which jointly processes camera and LiDAR data to achieve more robust and accurate object detection. This work addresses safety-critical applications in robotics and autonomous driving by fusing complementary sensor strengths. Casser's research consistently pushes the boundaries of learning from limited supervision, reducing reliance on costly labeled data while improving real-world performance. His contributions are foundational for developing perception systems that are both cheaper and more reliable, positioning him as a key innovator in scalable autonomous perception.