What’s the implementation of GRU cell in tensorflow?

We can use a chart to demonstrate the GRU cell implementation in Tensorflow, and let’s take a two cells GRU for example:

How difference

For multi-class classification, if you want optimize only one category during training, you should use SOFTMAX cross entropy. Otherwise, if you want to optimize more than one category, you should use SIGMOID cross entropy.

There is my implementation of the Large Margin Softmax Loss in TF.

This method could augment images by disturbing the RGB value of a image randomly.

Method

This method augment image by changing image through affine transformation.

The evolution of the Activation Style in Residual Blocks

The Activation Style is different between different residual structures.

Let’s see a very simple handwriting formula derivation

Define

Firstly, let define some variables and operations

namedtuple

If you are too lazy to create a class but you still want to use a variable that can act as a class object, then you should use namedtuple:

1
2
3
4

from collections import namedtuple
user = namedtuple('u', [1, 2, 3])
ztq = user(a='a', b='b',c='c')
print(ztq.a, ztq.b, ztq.c)

result:

1

(1,2,3)

Method

This method augment image by disturbing the YUV field of the target image and then transform back to RGB field.

The relationship with Densenet

Since I use shufflent_V2 on the task of Face Antispoofing for quit a time, I found that shufflenet_V2　is efficient yet also could provide high accuracy.