CODEWITHSUNDEEP
deep-learningconv-neural-networkpytorch
Jun
Jun

Reputation: 145

How can we calculate receptive field of a network includes transposed convolutional layer?

A network I designed includes transposed convolutional layer. (ConvTranspose2d in pytorch)

I want to get receptive field size of my network.

Does the concept of receptive field also hold on with transposed convolutional layer?

If yes, then how can I get it?

Upvotes: 6

Views: 4259

Answers (2)

Newbie
Newbie

Reputation: 67

You can use torch-scan to automatically compute receptive fields for a lot of layers. It provides support for those layers which are not supported by pytorch-receptive-field as well, such as Conv1D, ConvTranspose1D, etc. It also optionally gives effective stride and effective padding as well.

Upvotes: 3

Toyo
Toyo

Reputation: 751

You can use the library pytorch-receptive-field to automatically compute all layers' receptive fields.

Upvotes: 3

Related Questions