Projects

Build a lane change decision model with Compact Support Neural Network

One of the biggest obstacles in deploying Autonomous Vehicle (AV) is that without vehicle-to-vehicle communication, AV has to be able to predict human driver’s intentions and planning and control its action accordingly, while equally importantly, AV also has to behave just like a typical human driver such that other road users can infer the AV’s intentions. We propose to learn the human driver’s mental model in lane change decision making.

Playing Atari games with Hierarchical Reinforcement Learning

With options (i.e., multiple step actions) learned from manual book, hierarchical reinforcement learning achieve human-level learning efficiency and state-of-the-art performance in playing challenge games, e.g., Tennis, Pong, etc.

Scene Text Recognition

Detect, extract, and recognise texts in advertisement pictures. The recognized text will be feed to advertisement conversion prediction machine learning models.

Hua Huang

Projects

Build a lane change decision model with Compact Support Neural Network

Playing Atari games with Hierarchical Reinforcement Learning

Scene Text Recognition