This is a classical classification network for 1000 classes trained on ImageNet. The difference is that most convolutional layers were replaced by binary once that can be implemented as XNOR+POPCOUN operations. Only input, final and shortcut layers were kept as FP32, all the rest convolutional layers are replaced by BinaryConvolution layers.
|fp32 conv MFlops||960|
|bin conv MI1ops||7218|
The quality metrics calculated on ImageNet validation dataset is 70.69% accuracy
|Accuracy top-1 (ImageNet)||70.69%|
A blob with a BGR image in the format: [B, C=3, H=224, W=224], where:
It is supposed that input is BGR in 0..255 range
The output is a blob with the shape [B, C=1000].
[*] Other names and brands may be claimed as the property of others.