Sai-Kit Yeung

About

Sai-Kit Yeung is a multidisciplinary researcher whose work spans computer vision, robotics, 3D scene understanding, and simultaneous localization and mapping (SLAM). He is perhaps best known for introducing SceneNN (2016), a richly annotated RGB-D scene mesh dataset that addressed critical gaps in fine-grained labeling for computer vision and robotics research, amassing over 346 citations and becoming a foundational benchmark in the field. His contributions to real-time 3D semantic segmentation have further advanced the capabilities of autonomous systems—including drones and assistive robots—enabling on-the-fly dense scene reconstruction in indoor environments. More recently, Yeung has pushed the frontier of photorealistic mapping with Photo-SLAM (2024), a resource-efficient neural rendering and SLAM framework suitable for portable devices, already attracting over 151 citations within its first year. His research portfolio also reflects a breadth that few achieve, encompassing soft lattice structure design for additive manufacturing, object affordance reasoning, underwater visual SLAM, and early pioneering work in high-resolution tactile sensing and robotic object recognition using pseudorandom encoding. Collectively, Yeung's contributions demonstrate a sustained commitment to bridging perception, 3D representation, and physical interaction in intelligent systems.

Research Focus

Key Achievements

Top Papers

1. 1
   SceneNN: A Scene Meshes Dataset with aNNotations

   346 citations · 2016
2. 2
   Photo-SLAM: Real-Time Simultaneous Localization and Photorealistic Mapping for Monocular, Stereo, and RGB-D Cameras

   151 citations · 2024
3. 3
   Digital design and nonlinear simulation for additive manufacturing of soft lattice structures

   105 citations · 2018
4. 4
   Real-Time Progressive 3D Semantic Segmentation for Indoor Scenes

   79 citations · 2019
5. 5
   High sampling resolution tactile sensor for object recognition

   30 citations · 2002
6. 6
   Fill and Transfer: A Simple Physics-Based Approach for Containability Reasoning

   27 citations · 2015
7. 7
   Visual Object Recognition Using Pseudo-random Grid Encoding

   24 citations · 2005
8. 8
   Robotic Tactile Recognition of Pseudorandom Encoded Objects

   23 citations · 2004
9. 9
   Real-time GAN-based image enhancement for robust underwater monocular SLAM

   21 citations · 2023
10. 10
    Real-time Progressive 3D Semantic Segmentation for Indoor Scene

    9 citations · 2018

Key Collaborators

Contact & Links