search

forresti / SqueezeNet

SqueezeNet: AlexNet-level accuracy with 50x fewer parameters
BSD 2-Clause "Simplified" License
2.17k stars 723 forks source link

Some minor mistakes in the paper #65

Open Alphacch opened 11 months ago

Alphacch commented 11 months ago

Hello, thank you very much for your work. I would like to point out a small bug I found to you, there is a little problem with the first layer convolution of your network parameter picture. It wasn't until I checked your github that I discovered that it was a careless mistake in your paper. Before that, I spent a lot of time calculating the reason... 4518c0f553475365c5125fa56671965 ca016af4ccf68bc03e55bc3a6344974

Alphacch commented 11 months ago

Before this, I analyzed whether there was a problem with the results or the convolution parameters. Or whether there are any subtle differences in using different frameworks, etc.

forresti commented 11 months ago

I’m sorry about that! Glad you were able to figure it out. Good detective work. :)

Alphacch commented 11 months ago

https://github.com/forresti/SqueezeNet/issues/65#issuecomment-1711853315 thank you for your reply. Of course I've been using your squeezenet and I think it's great! Because caffe has stopped maintenance, I recently chose pytorch to build the network, so I discovered your problem. Thanks again for getting back to me.