Introduction to Convolutional Neural Networks for Visual Recognition

• A brief history of computer vision
  Organic vision > mechanical vision (first developed in 1545)

• recommend [VISION] - David Marr who is vision scientist
  - Stages of Visual Representation
  

• Normalized Cut (Shi&Malik) : means image segmentation (divide bulkly)
  Face Detection, SIFT(to match entire object to anoth object(seems like), Spatial Pyramicd Matching(devide part and use Support Vector Machine) and so on.
  

• IMAGE NET : has 22K categoris and 14M images
  In the Image Classification Challenge error rate was extreamly decreased by using CNN(AlexNet).
  

about CS231n

• focuses on one of the most important problems of visual recognition - img classification such as object detection, image captioning, action classification...

• CNN(Convolutional Neural Networks) is main tool for object recognition

• It was invented in 1998. However, there were not supportable things which make CNN more effective.
  In 2012, CNN finally shows its great performance by using more developed computers and big data.