MCS Capstone Projects

Year 4 Students of Yale-NUS College


The Yale-NUS curriculum culminates in an original capstone project which all students undertake with the guidance of Yale-NUS faculty and other subject matter experts. In the capstone program, students hone the disciplinary and general intellectual skills necessary to conceive, design and execute a year-long, self-directed project within their major. The program requires every student to engage with research in their discipline, to produce scholarship across appropriate formats, and to communicate their results to a variety of audiences. By completing their capstone work, students demonstrate independence, creativity and critical analysis.

Particularly, if you are interested in the areas of computer vision and machine learning/deep learning (applied to computer vision), please contact me as your potential supervisor. For general information on the projects, please see the recommended topics below.

Supervisor: Robby T. Tan (robby.tan [att]
Supervisor guidelines and regulations: PDF

Recommended Topics:

Generally I'm interested in applying deep learning techniques to computer vision problems, particularly to the problems of bad weather, motion analysis, and human pose/action recognition. However, if you have a specific plan in mind and you think my expertise can help, I am open for discussion.

  1. Visibility in Bad Weather
    Visibility can be degraded significantly by atmospheric particles such as fog or haze, rain streaks and raindrops. This can cause many computer vision algorithms to fail, since they usually assume a clear day scene. In this topic, it would be interesting to see if deep learning can help restore the visibility of degraded images due to bad weather. A few researchers have shown that Convolutional Neural Networks can be effective dealing with some problems (such as haze or sparse raindrops), however there are other bad weather conditions, where the problems are not yet explored. More information about this topic can be found here: Visibility in Bad Weather
  2. Motion Analysis
    Motion is one of the most basic cues of visual recognition. We can identify the presence of an object immediately after we notice its motion. Moreover, many high-level computer vision algorithms, such as, action classification or obstacle detection, rely heavily on motion information. One type of motion information is optical flow. While it has been explored for many decades, most of the optical flow algorithms are not robust to some conditions where the brightness constraint assumption is violated. Besides optical flow, problems in dynamic texture analysis is interesting to explore. Applications such as face expression or human action recognition will greatly benefit from robust motion extraction. More information about this topic can be found here: Motion Analysis
  3. Human Pose/Action Recognition
    Humans are central to many computer applications, such as, human-computer interaction, video surveillance, patients/elderly monitoring systems, store-goers behavior analytical systems, etc. Despite significant efforts to automate human action or pose recognition from video, the problem is not yet solved. This is due to the complexity of human actions and also due to the complexity of the surrounding world. Occlusions, intractable body localization, dressing styles, cluttered background etc. are part of the complexity. Solving some problems in this area will be interesting from both theoretical and practical perspectives. See the student projects below for some examples.

Selected Past Student Projects:

  1. Elena Ursu, "Pose Estimation in Video", 2013: [project page]
  2. Manuela Ichim, "Human Tracking and Orientation Estimation", 2013: [project page]
  3. Jeffrey Resodikromo, "Markerless 3D pose estimation", 2012: [project page]
  4. Michael Hobbel, "3D face reconstruction from a single image", 2012: [project page]
  5. Pascal Mettes, "Water Segmentation and Classification", 2013: [project page]
  6. Bob Boggemann, "Stereo Depth Estimation from Video Sequence", 2010: [project page]
  7. Charis Kontaxis, "Fluid Simulation", 2012: [project page]
  8. Timothy Kol, "Analytical Sky Simulation", 2012: [project page]
  9. Zhi Kang Shao, "Physically based rendering of mist and fog using MCML", 2013: [project page]