Computer Vision main
Main:
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