A guide to convolution arithmetic for deep learning
This provides a practical guide for deep learning practitioners to better design and analyze CNN architectures, but it is incremental as it synthesizes existing knowledge without introducing new methods.
The authors tackled the problem of understanding and manipulating convolutional neural network architectures by clarifying the relationships between properties like input shape, kernel shape, and output shape for various layer types, providing derived relationships and intuitive illustrations.
We introduce a guide to help deep learning practitioners understand and manipulate convolutional neural network architectures. The guide clarifies the relationship between various properties (input shape, kernel shape, zero padding, strides and output shape) of convolutional, pooling and transposed convolutional layers, as well as the relationship between convolutional and transposed convolutional layers. Relationships are derived for various cases, and are illustrated in order to make them intuitive.