search

forresti / SqueezeNet

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

Wrong Numbers? #6

Closed tmatas closed 8 years ago

tmatas commented 8 years ago

Hello, I'm trying to use this model, but the number in the various strata looks weird. Shouldn't the input be 3X224x224, as in the paper, instead of 3X227x227? And what does it mean that the first dimension is 10?

btw, awesome work.

forresti commented 8 years ago

Thanks for your kind words.

Good point about 224 vs 227. This is a typo in the paper. But, we have found that the accuracy of our CNNs aren't too sensitive to 224 vs 227. Substantially larger input image sizes might lead to higher accuracy.

In deploy.prototxt, 10 is the batch size. On Apr 18, 2016 1:28 AM, "tmatas" notifications@github.com wrote:

Hello, I'm trying to use this model, but the number in the various strata looks weird. Shouldn't the input be 3X224x224, as in the paper, instead of 3X227x227? And what does it mean that the first dimension is 10?

btw, awesome discovery.

— 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/6