Our new 3DOF pedestrian trajectory dataset is released ~ [link]
      The implementation of our latest research will be released soon~ [link]
      Gaussian Process for multi-class classification toolbox is published~ [link]
      Our latest high-resolution RGBD dataset for clothes recognition is published~ [link]
Li Sun (Kevin)    
PhD of  Computer Science.

I am a Research Fellow in L-CAS (Lincoln Center of Autonomous Systems), Univerisity of Lincoln. I am IEEE, BMVA, EUcog, SICSA member. 

Before I joined L-CAS, I was working as a Research Fellow in Extreme Robotics Lab, Univerisity of Birmingham.

My research focuses on the core challenges in the emerging robot vision to enable the robot to manipulate with complex industrial objects or drive in the dynamic, real-life environment e.g. warehouse, campus.

Research Statement:       

Currently, I am working on H2020 project ILIAD with Prof. Tom Duckett. This project aims to advance the state-of-the-art autonomous robot operation in a warehouse environment shared with humans. I am working on 3D Lidar-based perception, (i.e. semantic understanding, object detection), 2D/3D mapping in dynamic environment and autonomous navigation.

In the University of Birmingham, I was working on H2020 project RoMaNs with Prof. Rustam Stolkin. I am leading the computer vision research in this project, including 3D reconstruction, deep learning based 3D object detection and recognition, 3D tracking/pose estimation, etc. 

The objective of my PhD research is to advance the state of the art in perception and manipulation of deformable objects(such as clothes). The essence of manipulation with clothes is recognising the configuration of deformable surface and learning strategy to change the state of the surface. In my PhD research, our self-designed stereo-vision system is used to produce high resolution and accurate depth data, afterwards, a geometry-based feature framework is proposed to fully understand the configurations of the clothing using rich visual representations, and supervised learning is adapted to learning the recognition models, and non-parameterised model (GP) and reinforcement learning are employed to learn the manipulating skills. We have included this proposed pipeline in a robotic testbed used to investigate autonomous perception and manipulation of textiles and garments(European FP7 Strategic Research ProjectCloPeMa)

My previous research interest is machine learning, especially on algorithm innovation of multiple instance learning, and applying multiple instance learning algorithms to ECG classification/analysis and Image classification.

Research Interests :                                                                                                                

Computer Vision: 3D vision, 3D or RGBD object detection/recognition, visually-guided object manipulation                    
Machine Learning: deep learning,  sequence learning, reinforcement learning, mutiple-instance learning

Supervisor: Prof. Tom Duckett
Previous Supervisor: Prof. Rustam Stolkin
PhD Supervisor: Dr J. Paul Siebert &  Dr Simon Rogers