Non-photographic Image Categorization System

1
Non-photographic Image Categorization System
Progress Report
By Wang Fei Supervisor Min-Yen Kan
2
New classification Scheme
3
Progress since last meeting

• Modified classification scheme as shown in the
  last slide.
• Finished coding all image feature extraction
  methods, i.e. circle/rectangle/corner recognition
  and various color difference calculations based
  on image color histograms.
• Analyzed testing results on the 5000 image
  dataset.
• Acquired larger training data sets.
• Currently working on the first level of
  classification, i.e. b/t photo and non-photo.
• Also trying to run tests using a different
  machine learner, i.e. SVM.

4
Thank you