SNN_Calibration icon indicating copy to clipboard operation
SNN_Calibration copied to clipboard

Pre-training ANN model

Open lanyx7 opened this issue 2 years ago • 5 comments

Hello, I have tested the ANN model of VGG16 without usebn provided by you on ImageNet dataset, but the accuracy is only 55.074%. Could you please update the pre-training ANN model?

lanyx7 avatar Aug 03 '22 03:08 lanyx7

Do you use max pooling or average pooling in VGG?

yhhhli avatar Aug 05 '22 16:08 yhhhli

I used avgpooling in VGG. I used the VGG model you provided from SNN_Calibration/ImageNet/models/vgg.py directly.

Do you use max pooling or average pooling in VGG?

lanyx7 avatar Aug 05 '22 17:08 lanyx7

Got it, I will double check the code, and update it in this weekend.

yhhhli avatar Aug 05 '22 17:08 yhhhli

Got it, I will double check the code, and update it in this weekend.

Thank you for your reply. I also have tested the ANN model of VGG16 with BN and avgpooling provided by you on ImageNet dataset, it achieves an expected accuracy. Furthermore, I also tested the two ResNet34 pre-trained models you provided, which also achieved the expected accuracy. I hope these test results will help you.

lanyx7 avatar Aug 05 '22 17:08 lanyx7

just uploaded a new framework and new checkpoints, feel free to report any issues.

yhhhli avatar Aug 07 '22 00:08 yhhhli