Image Classification
  • a core task of computer vision
  • refers to the task of extracting information classes from a multiband raster image

Image Classification - Challenges

  1. semantic gap - an image is a tensor of integers between [0, 255] (i.e. 800x600x3 3 channels RGB)
  2. viewpoint variation - all pixels change when the camera moves
  3. illumination - all pixels change when the light source changes
  4. background clutter - some pixels changes when the background of the subject changes
  5. occlusion - parts of the subject are covered
  6. deformation - the subject can take various forms
  7. intraclass variation - the subject contains many subclasses
  8. context - the environment may modify the subject