Background

Ph.D. in Perceptual Intelligence – Visual perception of materials

  • An EU-founded research and training network PRISM– Perceptual Representation of Illumination, Shape & Material. It unites nine leading academic and industrial partners from across Europe to understand how the brain represents the physical properties of objects, surfaces and lighting in the surrounding world.

 

  • Focus: Material Perception in general and its interaction with Light Perception.

 

  • Project 1

We developed a novel probing method to measure material perception in a purely visual way without involving semantic information (i.e. not using words) and in a painterly approach. To test the material probe, we integrated it in a interface inspired by a DJ’s mixing desk and conducted a matching experiment.

  • Project 2

We implemented the method developed in project 1 to systematically study how canonical lighting modes influence the perception of our canonical material modes. In this study, a matching experiment was conducted using the same interface developed in project 1. New material-dependent lighting effects were found.

  • Project 3

We investigated the material-dependent lighting effects, and found asymmetric perceptual confounds between materials and lightings. Three experiments were conducted. The first experiment further developed the probe developed in Project 1 to allow optical mixing of lighting modes in a matching task, focused on studying whether observers can match optically mixed lighting modes. In the second experiment, a 4-category discrimination task was conducted, focused on studying whether observers can simultaneously discriminate materials and lightings. The third experiment was conducted to compare and relate the two type of tasks in the previous two experiments.

  • Project 4

We investigated how global environment lightings could evoke certain material qualities by conducting two scaling experiments for a list of terms that are commonly used to describe materials. The first experiment focused on canonical lighting modes we previously implemented and the second experiment focused on generic environments. Based on the results of the first experiment, we were able to make predictions of the light effects on material appearance for generic light environment. The results of the second experiment validated the prediction we made.

In this project we proposed a brilliance metric based on a spheric harmonics analysis on the lighting environment.

  • Project 5

We further investigated how to optimize the environment lightings for certain material qualities, e.g. by rotating a globally optimized environments.

  • Results: We developed a novel material probe for quantitatively and purely visually measuring the visual perception of materials; created dataset of controlled stimuli varying parametrically in canonical material and light modes. ;/
  • Visit Experience:
    • Feb.15 – May.15, University of Giessen, Germany. Hosted by Roland Fleming.
      • 3D modeling and rendering for the canonical material modes in our probe (part of project 1)
      • Results of MDS experiment for Glossy perception study collaborated with local car paint industries.
    • Apr.16 – May.16, INRIA – University of Bordeaux, France. Hosted by Pascal Barla
      • Gratin & OpenGLSL
      • project 4 & 5

M.Sc. in Robotics (Graduate with Distinction)

  • Core Courses
    • Artificial Intelligence
    • Computer-aided Manufacturing and Design,
    • Computer Vision
    • Pattern Recognition
    • Real-Time Systems and Control
    • Robotics Systems
    • Sensors and Actuators
  • Thesis: A Neural Network for Solving the Stereo Correspondence Problem.
    • I developed a stereovision version of an existing neural network model of primary visual cortex cells, implemented the model on simple artificial scenes and complex realistic scenes (MATLAB).

B.Eng. in Mechanical Engineering and Automation (ZH-CN/EN Bilingual program)

  • Thesis: The Identification of Tool Cutting Condition Based on AE (Acoustic Emission) Signal.
    • I helped building the hardware (for lathe machine & milling machine) and software (LabVIEW + MATLAB) platforms for receiving and processing the AE signal to identify specific tool conditions in manufacturing processes implementing pattern recognition method.