search

SongjieXie / Hyper-AJSCC

Pytorch implementation of "Deep Learning-Based Adaptive Joint Source-Channel Coding using Hypernetworks"
19 stars 4 forks source link

Questions about compression rates #1

Open Luckwjf opened 2 months ago

Luckwjf commented 2 months ago

Hi teacher, I would like to ask why num_symbols_real_img = args.sc * 2 in the code is multiplied by 2 here?

SongjieXie commented 2 months ago

As the comments above this line of code explained, the encoded features will directly transmit the real and imaginary parts of the channel input. Thus, the dimension of features (num_symbols_real_img) is two times the dimension of channel input (args.sc).