CODEWITHSUNDEEP
tensorflow
dtracers
dtracers

Reputation: 1658

Reduce shape of tensorflow model

So I am creating a tensorflow model but it uses an existing model whose output has a shape of [1, 512] or something like that. I want to add a couple of layers to it at the end but it needs to end up with the same output shape as the number of classes which is something much smaller

[1, 5] for example.

What is the best way to reduce the output so it matches without losing too much data?

I tried pooling but I don't think I understand it enough.

Upvotes: 0

Views: 66

Answers (1)

Vijay Mariappan
Vijay Mariappan

Reputation: 17201

You can connect the output to a dense layer like: logits = tf.layers.dense(outputs, units=5) to reduce from 512 to 5 output units.

Upvotes: 1

Related Questions