Teaching

Lecture

  • 2019/20 [06-25024] Robot Vision
  • 2018/19 [06-25024] Robot vision

Final Year & MSc Project

Dr. Hyung Jin Chang h.j.chang@bham.ac.uk Office hours (Online): Monday 2-4 pm (Join Zoom Meeting: https://bham-ac-uk.zoom.us/j/81104733867?pwd=ZkJCSHZrRjNncFQxa2Y1NmpxOVRoZz09 Meeting ID: 811 0473 3867 Passcode: 179454) —

My research interests are focused on human-centred visual recognition and understanding, especially in application to human-robot interaction. Computer vision and machine learning including deep learning are my expertise areas. Most of the projects are going to be interesting and fun to perform because you will have visual results to enjoy and experience “deep learning” techniques. I expect the students to have intermediate understandings of computer vision and machine learning. If you have strong coding skills in Python, Matlab or C++, then it can be extended to “advanced” and “challenging” topics. Highly suggested topics for 2021:

  • Efficient quantization techniques for deep learning
  • Appearance-based pedestrian attribute recognition (gender, height, weight, glasses, clothes etc).
  • Scene context-based human-object detection

Topic introduction with slides in a video (used on 27th Feb 2019): link

Below are some general research project areas:

  • Real-time gaze estimation
  • Real-time hand pose estimation
  • Real-time head pose estimation
  • Human body pose estimation
  • Human facial expression recognition
  • Inference on the visual attention of deep network
  • Visual object tracking
  • Egocentric vision
  • Autonomous vehicle related (Monitoring driver and passengers)
  • Transfer learning
  • Multi-task learning

Also, I am interested in many other computer vision and machine learning problems.

Here are more specific potential project topics for this year (2020-2021):

  • Tablet camera focusing point guidance by eye gaze
  • Eye gaze and language-based region selection system
  • Efficient eye gaze network design
  • Eye gaze-based wheelchair control system
  • Student attention level monitoring system
  • Body sonification system
  • Hand gesture recognition
  • Visual object tracking
  • People counting system
  • Human face tracking system on a mobile phone (https://github.com/google/mediapipe)
  • 6D object pose tracking system
  • Yoga coach app
  • Golf swing coaching app
  • Glasses-wearing detection network (Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts)
  • Wine recommendation system via artistic painting
  • and many others.

All my research assignments will use deep learning and should be based on an understanding of basic computer vision. If you’re confident in deep learning (especially CNN) and think you’re good at implementing, challenge it!

I am happy to supervise your own enthusiastic ideas if those are relevant to computer vision and machine learning tasks.