resnet50-binary-0001

Use Case and High-Level Description

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.

Specification

Metric

Value

Image size

224x224

fp32 conv MFlops

960

bin conv MI1ops

7218

Source framework

PyTorch*

Accuracy

The quality metrics calculated on ImageNet validation dataset is 70.69% accuracy

Metric

Value

Accuracy top-1 (ImageNet)

70.69%

Inputs

A blob with a BGR image and the shape 1, 3, 224, 224 in the format B, C, H, W, where:

  • B – batch size

  • C – number of channels

  • H – image height

  • W – image width

It is supposed that input is BGR in 0..255 range

Outputs

The output is a blob with the shape 1, 1000 in the format B, C, where:

  • B - batch size

  • C - predicted probabilities for each class in [0, 1] range

Use Case and High-Level Description

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.

Specification

Metric

Value

Image size

224x224

fp32 conv MFlops

960

bin conv MI1ops

7218

Source framework

PyTorch*

Accuracy

The quality metrics calculated on ImageNet validation dataset is 70.69% accuracy

Metric

Value

Accuracy top-1 (ImageNet)

70.69%

Inputs

A blob with a BGR image and the shape 1, 3, 224, 224 in the format B, C, H, W, where:

  • B – batch size

  • C – number of channels

  • H – image height

  • W – image width

It is supposed that input is BGR in 0..255 range

Outputs

The output is a blob with the shape 1, 1000 in the format B, C, where:

  • B - batch size

  • C - predicted probabilities for each class in [0, 1] range

Legal Information

[*] Other names and brands may be claimed as the property of others.