Hi everyone,
How to perform full convolution in tensorflow , this method is explained in this article ( How does Backpropagation work in a CNN? | Medium ) . the function tf.nn.conv3d has parametre padding which accepte two types "SAME"and “VALID” but there is not “FULL”
thanks .
1 Like
Hi @Abderrezzak_FELIOUNE, Could you please explain what the full padding means? How does the padding be applied in the full padding? Thank you.
Hi @Kiran_Sai_Ramineni ,
the full convolution is explained in this article
Thank you.
Hi @Abderrezzak_FELIOUNE, I have gone through the document you have shared regarding the full convolution. At present there is such a type of convolution present in the Tensorflow. The convolution in Tensorflow will be performed by placing the top left filter on the top left value of the input.
But as an alternative you can achieve this by adding one extra padding to the input when the same
value is passed to the padding argument in the convolution or you have to add 2 padding layers if valid
is passed to the padding argument in the convolution.
Note: The padding size will be dependent on the size of the kernel.
please refer to this gist for working code example. Thank You.
Hi @Kiran_Sai_Ramineni ,
thank you for your response it is very clear ,you are right we have to implemente by padding and convolution with “VALID”