Allan Raventos

Papers

Total Citations

H-Index

About

Allan Raventos is a researcher in computer vision and robotics, with a primary focus on self-supervised 3D perception for autonomous systems. His most notable contribution is the development of PackNet, a novel deep network architecture for monocular depth estimation that learns to infer 3D structure solely from unlabeled video streams. This work, published in 2020 and cited 35 times, addresses a critical challenge in robotics: enabling 3D perception from ubiquitous, low-cost cameras without the need for expensive active sensors like LiDAR. By combining geometric principles with a carefully designed 3D packing and unpacking scheme, Raventos’s method achieves state-of-the-art performance in self-supervised depth estimation, paving the way for more accessible and scalable robotic perception. His research has significant implications for autonomous navigation, drone flight, and augmented reality, where robust depth sensing is essential but hardware constraints often limit sensor choice. Raventos’s work stands out for its practical elegance—turning a fundamental limitation (monocular video) into a powerful training signal—and represents a meaningful step toward truly self-sufficient visual intelligence in the wild.