3
$\begingroup$

I have a problem where I wish to classify each column in an image of a feature being present (1) or absent (0) in each column. The output of the model should be a vector of size of the width of an image. I am trying to consider how the output of a CNN would look for such a network. I assume I would use a sigmoid loss function due to being a binary classification task. I can see how it would be possible to output a single 1 or 0 from the output of the sigmod function. However, I cannot seem to figure out how I would be able classify each column in the image. Would anyone have any ideas on how to implement this? I believe I do not have the correct idea of the fully connected layer and/or sigmoid function. Would appreciate any help in understanding this.

$\endgroup$
2
$\begingroup$

The last layer of the NN will contain as many neurons as the number of columns. And, they can use sigmoid activations (not softmax because the output will probably consist of multiple 1s).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.