Computer Vision main


Main:

CNN


Def:

  • process understand visual data
  • visual processing starts with simple structure

visual data:

how visual is physically formed:

neural network, convolutional neural network:


Process:

input image -> Primal sketch -> 2 1/2 -d sketch -> 3-D model

  • Input image:
    • Perceived intensities
  • Primal Sketch
    • Zero, crossings, blobs, edges, bars
  • 2 1/2 -d sketch
    • Local surface orientation
  • 3D model representation
    • 3-D models hierarchically organized

Recognize and Represent objects

goal: reduce the complex of the complex structure of the object into a collection and their geometric configuration

Method:

  • generalized Cylinder

  • Pictorial Structure Object recognition


image classification:

Image segmentation or semantic segmentation is the task of assigning a class to each pixel in an image. Models are trained using segmentation maps as target variable. z


TAGS