Description

This course covers advanced topics in computer vision in which vision problems are formulated and solved by inference from noisy and uncertain data from statistical learning viewpoint. Topics include learning algorithms and their applications to computer vision problems, as well future research directions.

Lectures

·        CSIE building 105

·        6:00 to 8:45 pm

Office Hours and contact information

  • Instructor: Ming-Hsuan Yang (Please call me Ming-Hsuan)
  • Email: mhyang @ csie.ntu.edu.tw
  • Students are encouraged to send me emails for questions and project discussions
  • Office hours: whenever I am at office

Topics

  • Dimensionality reduction (2.5 lectures, 09/28, 09/29, 10/03):  principal component analysis, factor analysis, probabilistic principal component analysis, mixture of probabilistic principal component analyzers, mixture of factor analyzers, isomap, locally linear embedding.
  • Classifier (1.5 lectures, 10/03, 10/04): Fisher linear discriminant, support vector machine, relevance vector machine, kernel machines, adaboost.
  • Generative model (2 lectures, 10/19, 10/20): graphical model, Bayesian inference, belief propagation, Gaussian process, EM algorithm.
  • Approximate inference (2 lectures, 10/24, 10/25): Markov chain Monte Carlo, variational learning.
  • Visual tracking (2 lectures, 11/21, 11/22):  particle filter, mean shift, 2D/3D human tracking.
  • Dynamics (2 lectures, 11/23, 11/24): Autoregressive models, linear dynamic system, Kalman filter, dynamic textures, video synthesis.
  • Image feature (1 lecture, 12/26): interest point, SIFT, exemplar.
  • Object detection (1 lecture, 12/27): face/car/pedestrian detection, human pose estimation.
  • Other topics (1 lecture, 12/28): Markov random field, conditional random field, convex optimization (to be determined).
  • Project presentations (1 or 2 lectures, 12/29, 12/30).
  • Check the course web page for most recent update

Format

  • Project oriented course
  • 3 to 4 lectures monthly (each one lasts for 3 hours)
  • Meeting time to be determined
  • Students are encouraged to ask questions (i.e., class participation)
  • No exams or quizzes
  • Lectures will be given in English

 

Reading list

  • Reference for background study: 
    1. Computer Vision: A Modern Approach, David Forsyth and Jean Ponce. Prentice Hall
    2. Neural Networks for Pattern Recognition, Christopher Bishop, Oxford University Press
    3. Pattern Classification, Richard Duda, Peter Hart, and David Stork, Wiley Interscience
    4. The Elements of Statistical Learning: Data Miming, Inference and Prediction, Trevor Hastie, Robert Tibshirani and Jerome Friedman, Springer
  • Readings will be from the text and additional material that will be handed out or made available on the web page. 
  • All lecture slides will be available on the course website.

Prerequisites

  • Basic knowledge in computer vision, probability and statistics
  • Programming languages: Matlab and others (e.g., C++)

Requirements

  • Literature review and critique: students are expected to read conference/journal papers and submit critiques weekly.
  • Term project: students are expected to work on a term project individually or in two-member groups.
  • Oral presentations: students need to make one presentation on project overview and progress, as well as one final presentation in English.
  • Final project report: students are required to submit a project report.
  • Formats and details regarding all the abovementioned items will be available on the course web site.

Grading

  • 10% Class participation
  • 20% Critiques
  • 20% Oral presentations
  • 50% Project