Which model is used to calculate the total parameters in paper table 2? #5
Description
Hi, @tyshiwo , Thanks for your excellent work.
I have read the MemNet paper, and I find that in table 2 the Memnet parameters is 677K, I'm confused by the total parameters.....
The default model is M6R6 with depth 80. In every residual block, there are 2 relu_conv structure (64x64x9x2), so I just calculate the conv parameters:
(64x64x9x2)x6x6 = 2654208
and I calculate the total parameters, which are roughly 2.91M.
So I'm confused by the total parameters. Can you point out the actual method to calculate the parameters in your paper? Thanks a lot.