How it works

1. Convolution operation

• convolution operations can make the detector dependent on features and independent of locations -> coarse and invariant detection
• stride = the moving step size of the feature detector: usually 2
• feature detector = kernel = filter :usually 3x3, but other models such as AlexNet use 7x7
  

2. Pooling

• Pooling(=down-sampling) can reduces features and is more robust to overfitting.
• types
  • max pooling: represent a subregion by max activation map
  • global pooling: reduce spatial dimension to 1
  • aum pooling, Avg Pooling(~subsampling)
    

3. Flattening

4. Full connection

The flattened vectors are propagated into an ANN as the input layer