Discovering the patterns to enable machines to recognise, understand and analyse human behaviour.

Headshot of AAII's Professor Xiaojun Chang
Professor Xiaojun Chang | Lab Director

Computer vision and machine intelligence

The Recognition, Learning and Reasoning (ReLER) Lab is committed to enable machines to accurately recognize the environment, adaptively understand the human interactions, and autonomously analyse behaviour through reasoning. To this end, we work on computer vision, learning algorithms, natural language, and their intersections.  Concretely, we aim at developing novel methods for object, face, action, and event recognition by localising positions and segmenting the instances in images and videos.

Besides recognition, it is also essential for the machines to communicate fluently with humans by understanding natural language instructions and queries. To bridge the literacy gap, ReLER Lab develops captioning, question answering and dialogue systems for better visual understanding and reasoning.

Upper L: City scene with outline below in different colours showing people and objects such as bikes. Parsing the scene into different semantic regions. Upper R: People on steps moving or sitting, incl.a pickpocket.Identifying semantic concepts in large-scale image data. Lower left: ReLER: Recognition (cat), Learning (bulb), and Reasoning (machine Learning icon). Lower R: Photo of ballet dancer in different positions. Identifying the instance-level actions related to given language expresssion in videos.

Upper left: Parsing the scene into different semantic regions. Image: Cityscapes.
Upper right: Identifying semantic concepts in large-scale image data. Image: Data to Decision CRC.
Lower left: ReLER: Recognition, Learning, and Reasoning.
Lower right: Identifying the instance-level actions related to given language expresssion in videos.