识别问题的综述框架

Recognition, reconstruction and reorganization。
Recognition is about attaching semantic category labels to objects and scenes as well as to events and activities.
识别问题就是贴标签。

In the early days of computational vision, in the 1970s and 1980s,
there was a broad agreement that vision could be usefully broken
up into the stages of low level, mid level and high level vision.
[58] is perhaps the best articulation of this point of view, with low
level vision corresponding to processes such as edge detection, mid
level vision leading to representation of surfaces, and high level vi-
sion corresponding to object recognition.

低级是边缘 中级是轮廓 高级是物体，这是一个自下而上的过程，但在后来被证实infeasible