CODEWITHSUNDEEP
tensorflow
P-Gn
P-Gn

Reputation: 24641

Is there a systematic way to compute the receptive field of a neuron?

I am interested in computing the receptive field of a neuron relatively to the input, or more generally relatively to an earlier layer.

This can be done manually, but I would like to know if there is a built-in function to do it or otherwise if there is a way do compute it automatically.

Is there something that could work at least on a simple (single stream, no skip/concatenation) network restricted to convolutions and reduce layers, with possibly a mix of SAME and VALID paddings and non-unit strides?

Upvotes: 1

Views: 316

Answers (1)

André Araujo
André Araujo

Reputation: 86

Yes, as of Aug 2017, you can simply use tf.contrib.receptive_field

See https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/receptive_field for details.

Upvotes: 1

Related Questions