Closed yassouali closed 4 years ago
Ok, sorry, I should have checked the appendix first, so there is two possible ways of computing the negative and positive samples, either (1) by using a discriminator or (2) by computing a non linear embedding of the local and global feature maps with the possibility to evaluating a large number of positives and negatives. And the implementation above is for the second case.
The details are in appendix A2.
First of all, thank you very much for providing the source code for your work,
I was going though the code to get a better understanding of the paper, and I have some questions, So for CIFAR 10 in case of Local DIM, the networks are as follows:
The encoder
Global network, taking as input the output of the encoder and outputing the global feature vector of shape [64, 2048, 1, 1]
And the local network, with the same architecture as Global net, taking as input the output of layer 1 of the encoder (shape of [64, 128, 8, 8]) and output the local feature of shape [64, 2048, 8, 8] reshaped to [64, 2048, 64].
With local and global features, the next step of to compute the loss by computing the product of the two features in the form of [64, 64, 64, 1] and taking the positives from the diagonal and negatives as the rest.
My question, is what part of the network constitues the discriminator, given that the fake and real samples are computed as follows:
![image](https://user-images.githubusercontent.com/45082182/73274732-54403900-41e6-11ea-8f26-c3b096a7cf18.png)
I can't see where does the Discriminator
Dw
appears in the implementation. Thank you.