search

forresti / SqueezeNet

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

About the training data resolution #8

Closed austingg closed 8 years ago

austingg commented 8 years ago

Hi, I have a question about the training data resolution of SqueezeNet. Is it resized to 256x256 or with the smaller side 256 px. I have use both resolutions to test your pretrained model. Seem 256x256 resolution get a higher accuracy.

forresti commented 8 years ago

I believe our training data is stretched so that is square 256x256. There are certainly alternatives to this that will lead to higher accuracy (though if you change the input dimensions a lot, you may want to retrain). On Apr 27, 2016 8:37 PM, "Yubin Wang" notifications@github.com wrote:

Hi, I have a question about the training data resolution of SqueezeNet. Is it resized to 256x256 or with the smaller side 256 px. I have use both resolutions to test your pretrained model. Seem 256x256 resolution get a higher resolution.

— You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub https://github.com/DeepScale/SqueezeNet/issues/8

austingg commented 8 years ago

@forresti got it, thanks.