Different kernel size from paper

Xiaoccer / MobileFaceNet_Pytorch

MobileFaceNets: Efficient CNNs for Accurate Real-Time Face Verification on Mobile Devices

348 stars 89 forks source link

Different kernel size from paper #13

Open quocnhat opened 4 years ago

quocnhat commented 4 years ago

Hello, Thanks for sharing this project. I find that you use kernel size of (7x6) at layer linear7, while it is (7x7) in the paper. Why is this change? Thank you

cavalleria commented 3 years ago

when input is 112x112 , the output is 7x7, and input is 112x96, the output is 7x6