search

GANs-in-Action / gans-in-action

Companion repository to GANs in Action: Deep learning with Generative Adversarial Networks
1.01k stars 420 forks source link

need clarification for P.51 #1

Closed amiltonwong closed 5 years ago

amiltonwong commented 6 years ago

Hi, authors,

In page 51, it is mentioned as:

6. Use transposed convolution, transforming the 28x28x64 tensor into the output image
size, 28x28x1.

However, according to Fig.4.2, it should be the following:

6. Use transposed convolution, transforming the 14x14x64 tensor into the output image
size, 28x28x1.

Please clarify for this. THX!

vbok commented 5 years ago

Thank you for bringing this to our attention. The chapter has an error. The correct progression, as reflected in the diagram and the comments in the code, is the following:

Noise vector → 7x7x256 layer → 14x14x128 layer → 14x14x64 layer → 28x28x1 output

The correction should be reflected in the upcoming MEAP update to our book.

amiltonwong commented 5 years ago

@vbok ,

Thanks for your clarification.

BTW, when will the next chapters be released? I've read through the contents from ch1 to ch4.

vbok commented 5 years ago

The MEAP update with Chapter 5 should go live no later than Wednesday next week, possibly sooner. Subsequent new chapters will be added on a monthly basis.