"CS231N Winter 2016"

Lecture 1

cs131, undergraduate course in computer vision, Introduction.
cs231a, graduate program in computer vision, covers a wider range of CV knowledge (cs231n is primarily the vision Recognition/image classification) aspiring CV learners recommend learning
cs231n, mainly CNN for image classification

History is not a lot of nonsense. It is said that millions of years ago the birth of biological vision greatly facilitated the evolution process, because the predator has eyes will chase, the arrested Eaters have Eyes will run Blablabla ...

Harvard's two-man experiment with cats (Nobel Prize) found that when switching pictures, the cat's brain vision was excited, indicating that V1 was sensitive to edges and edges orientation.

1963, Larry Roberts that the edge is the key, it is the edge (even if the color or orientation change) determines the shape of the object. His PhD dissertation is about this, a pioneering article on CV.

、

But the official birth of the CV was supposed to be in 1966, when MIT AI Lab (1960 's early established two AI labs, one was Marvin Minsky was founded at MIT and the other was John McCarthy (McCarthy, A professor at Stanford, who is also the inventor of the word AI, felt that a summer should be used to solve the problem of computer vision. This is Li Feifei spit, and want to use a summer project to solve the CV problem.

2016 is the CV's 50 birthday!

another person to remember is David Marr ()
Huber&wiesel (1959, Harvard) tells us that the vision is very simple, is the edge;
David Marr (1970s) tells us that human vision systems should have different stages/layers, resulting in complex vision.

David Marr did not point out a mathematical model similar to that of the visual system, but its stages thought was eventually absorbed by Deep learning/cnn.

Since then, following the construction of the 3D vision, the first wave of vision algorithm wave has come. Several representative jobs were introduced:

Brooks (1979) thinks that three-dimensional objects are composed of simple objects (such as cylinders), and Fischler (1973) Think ... What does it mean?

In the following diagram, Fei Fei said it was a representative work at the time, the full text is to use a very simple edges and shapes to identify objects, to show that the previous CV is really primitive.

Since then, the 90 's finally entered the color picture era, at that time very influential work is not how recognize an object, but carve out images into sensible parts

In CV also become perceptual grouping, is a very basic problem, still not very good solution.

"CS231N Winter 2016"