About Joint Training

[DongfeiJi]

Thanks for your work, but I am wondering how to joint training, which is mentioned in the paper (Algorithm 1 Joint Learning Scheme).

[yanfjz]

I have two related questions as well:

1. Is the "moving state" S the deltas between src_pts and dst_pts (src_pts - dst_pts) which contains 2(N+1) points?
2. Is the loss function for joint learning cross-entropy loss? If so, what is the representation of GT for this loss function?
   Any explanation/idea/suggestion/discussion is welcomed :)

[Canjie-Luo]

@DongfeiJi
This version is the default random augmentation. We guess it is sufficient for practical use.

[Canjie-Luo]

@yanfjz

1. The "moving state" denotes the moving directions.
2. Yes, we use cross-entropy loss. The GT is the moving state that increases difficulty.

[zdz1997]

@Canjie-Luo
I read the paper Learn to Augment: Joint Data Augmentation and Network Optimization
for Text Recognition.
and I'm so interested in joint training (Algorithm 1 Joint Learning Scheme)，can I ask for the whole code of paper?There is something I can't understand and I want to retrain it .I'll appreciate it very much .Here's my email [email protected]

[DaeHwanGi]

@Canjie-Luo

I have some questions about the paper.

1. Output size of the agent network is 2x(N+1)x2x2. Isn't just two coordinates enough to predict the direction of each point?
2. In Algorithm 1, randomly select one point in S and switch to the opposite direction to make S prime. Does this mean choosing one moving state in a mini-batch?

[shubham303]

I read the paper Learn to Augment: Joint Data Augmentation and Network Optimization
for Text Recognition.
and I'm so interested in joint training (Algorithm 1 Joint Learning Scheme)，can I ask for the whole code of paper?
my email: [email protected]

[firatkizilirmakk]

@shubham303
Have you got the code?

I would be glad to get joint training code as well.
my email: [email protected]

Thanks in advance!

[matiascoronados]

@shubham303
I would really appreciate if you can share that code with me!

my email: [email protected]

Thanks :)

[shubham303]

@matiascoronados I don't have the code.

[Canjie-Luo]

Hey guys, due to the intellectual property protocol, I cannot release the code.