Silvio Savarese

About

Silvio Savarese is a prominent computer vision and robotics researcher whose work spans 6D object pose estimation, multi-object tracking, embodied AI, and autonomous navigation. Best known for foundational contributions at the intersection of perception and robot learning, Savarese has shaped how intelligent agents understand and interact with the physical world. His landmark paper *DenseFusion* (2019, 1,121 citations) introduced an influential framework for 6D object pose estimation from RGB-D data, becoming a benchmark reference in the field. Complementing this, *6-PACK* extended pose estimation to real-time category-level tracking, while early work on depth-encoded Hough voting demonstrated lasting influence in 3D scene understanding. His *Gibson Environment* (714 citations) addressed a critical gap in embodied AI by enabling agents to train on photorealistic real-world simulations, accelerating research in sensorimotor learning. Savarese has also made significant strides in trajectory forecasting and social robot navigation through *Social GAN* and *Social-BiGAT*, and advanced multi-object tracking with decision-making frameworks accumulating over 700 citations. His Neural Task Programming work further demonstrates a commitment to generalizable, hierarchical robot learning. Across these contributions, Savarese has helped define the modern agenda for intelligent, perception-driven robotic systems.

Research Focus

Key Achievements

Top Papers

1. 1
   DenseFusion: 6D Object Pose Estimation by Iterative Dense Fusion

   1,121 citations · 2019
2. 2
   Learning to Track: Online Multi-object Tracking by Decision Making

   716 citations · 2015
3. 3
   Gibson Env: Real-World Perception for Embodied Agents

   714 citations · 2018
4. 4
   Social-BiGAT: Multimodal Trajectory Forecasting using Bicycle-GAN and\n Graph Attention Networks

   278 citations · 2019
5. 5
   Scene Memory Transformer for Embodied Agents in Long-Horizon Tasks

   186 citations · 2019
6. 6
   Learning task-oriented grasping for tool manipulation from simulated self-supervision

   185 citations · 2019
7. 7
   Social GAN: Socially Acceptable Trajectories with Generative Adversarial Networks

   180 citations · 2018
8. 8
   6-PACK: Category-level 6D Pose Tracker with Anchor-Based Keypoints

   159 citations · 2020
9. 9
   Neural Task Programming: Learning to Generalize Across Hierarchical Tasks

   153 citations · 2018
10. 10
    Depth-Encoded Hough Voting for Joint Object Detection and Shape Recovery

    150 citations · 2010

Key Collaborators

Contact & Links