Does the original model have two softmax outputs?

Question

The V3 launch model is shown in this image:

Image from this blog post:

https://research.googleblog.com/2016/03/train-your-own-image-classifier-with.html

There seem to be two outputs to the Softmax classification. Why is this?

Which one is used in the TensorFlow example as an output tensor named "softmax: 0" in this file?

https://github.com/tensorflow/tensorflow/blob/master/tensorflow/models/image/imagenet/classify_image.py

In the academic work for the Inception v3 model, there is no such image of the Inception model:

http://arxiv.org/pdf/1512.00567v3.pdf

I am trying to understand why these two network branches exist with seemingly two different softmax outputs.

Thanks for any clarification!

Answer 1

4 , . , , , . , softmax:0 , , auxiliary_softmax:0.