How to define num of output in Softmax Layer

[Antonio Paes]

Hi guys, I start a train with 360.000 images for train, and 150.000 for validation, and your labels are 0, 1, -90, 45 , -35...... N.

I have 13 class, distrubuted for all images, and the images are RGB, size 54x54.

but in my validation tests accuracy no pass of 0.07823. Softmax layer have something to do with this bad accuracy?

If yes, how I calculate output os softmax layer?

Thanks.