About why two fully connected layers are generally used together - Gaaray - Blog Park
Why do some CNN networks have two fc fully connected layers at the end? - Know almostAbout why two fully connected layers are usually used together - Weixin_30376453 Blog - CSDN Blog
The 1x1 convolution can be summarized as follows:
1. In the middle of the network
1.1 The first layer 1x1 linearly combines the features extracted by the convolution layer, and the second layer realizes a highly nonlinear transformation of the input data;
1.2 Expand the model capacity, Increase generalization (same as increasing network depth);
2. At the end of the network
2.1, map the extracted features to the corresponding output scale for classification;
2.2 Combine the features extracted by CNN to adapt to their own data sets.
