24. CNN Architecture

A basic CNN can be constructed from:

Input
 ↓
Convolution
 ↓
ReLU
 ↓
Pooling
 ↓
Convolution
 ↓
ReLU
 ↓
Flatten
 ↓
Linear
 ↓
Output

pytorchcomponents include:

nn.Conv2d
nn.ReLU
nn.MaxPool2d
nn.Flatten
nn.Linear

CNNs learn hierarchical visual representations.