[深度学习论文笔记][Object Localization] OverFeat: Integrated Recognition, Localization and Detection using C

Sermanet, Pierre, et al. “Overfeat: Integrated recognition, localization and detection using convolutional networks.” arXiv preprint arXiv:1312.6229 (2013). (Citations: 978).

1 Motivation

Objects of interest sometimes vary significantly in size.

1.1 Sliding Window Method
Apply classification network over multiple locations and multiple scales on a image. However, many viewing windows may contain a perfectly identifiable portion of the object (say, the head of a dog), but not the entire object, nor even the center of the object. This leads to decent classification but poor localization. 

Thus, we train the system to not only produce a distribution over categories for each window, but also to produce a prediction of the location and size of the bounding box
containing the object relative to the window.

1.2 Classification and Regression Network
We run classification and reg