Projects

Reflection-aware visual understanding

Reflections are ubiquitous in our environment. Although provide only ‘subtle’ information, visual reflections bring flexibility to our visual perception of the environment. This project aims to explore the potential use of reflections in computer vision system. It will mainly address two questions: 1) will the presence of visual reflections affect the performance of current computer vision systems; 2) how to make effective use of visual reflections to help the reconstruction and understanding of the environment. Studying these questions will give new insights to computer vision systems and has potentials in industrial applications.

Vision-based 3D surface imaging

To reconstruct the profile of a 3D surface has wide applications in reverse engineering, computer-aided medical diagnosis, in-situ defect detection for manufacturing processes, etc. Vision-based solutions have the advantages of being low cost, non-destructive, and with easy implementation. This project aims to explore solutions integrating advances in imaging techniques, artificial intelligence, and robot manipulation for efficient, large scale, and high resolution 3D surface reconstruction.

Visual analytics for machine learning

Machine learning has seen great success in a wide range of applications. However, machine learning, especially deep learning, has most often been treated as a black box because of its unclear working mechanism. This project aims to open the black box by developing interactive visualisation and integrate it into the development of machine learning algorithms at various stages. It will assist the developers to understand the learning process, diagnose problems, and refine the machine learning models. It will also help produce explainable models to enable users to understand, trust, and manage Artificial Intelligence (AI) systems.

3D face modelling and processing

Faces play an important role in human social interactions, and are of special interest in photography and drawings. Our vision system has been well tuned to detect and recognise faces under different environments and from different portrait forms, e.g., photos, paintings, sketches, etc. Moreover, our vision system can easily model vivid 3D face shapes out of these 2D portrait forms. This project aims to investigate how to equip computer vision system with the same ability of 3D face modelling, and further assisting the tracking and recognition of faces, and face stylisation (e.g. caricaturization, bas-relief generation).

Interactive shape from shading

Shading variation on a surface indicates its 3D shape. Shape from shading (SFS) is a classic computer vision technique to recover the 3D shape from the 2D shading information. However, the reverse engineering from 2D to 3D is a non-unique process and has several kinds of ambiguities. In this project, we developed an interactive framework to tackle the intrinsic problems of SFS, and combined heuristic search to improve the intuitiveness of user interaction.

Page updated

Google Sites

Report abuse