MobileNetV2

class torch_ecg.models.MobileNetV2(in_channels: int, **config: torch_ecg.cfg.CFG)

Bases: torch.nn.modules.container.Sequential, torch_ecg.utils.utils_nn.SizeMixin, torch_ecg.utils.misc.CitationMixin

MobileNet V2.

MobileNet V2 is an upgraded version of MobileNet V1, originally proposed in 1. It uses inverted residual blocks instead of the original residual blocks. Torchvision’s implementation [#v2_pt] and Keras’ implementation 3 are used as references.

Parameters

• in_channels (int) – Number of channels in the input signal tensor.

• config (dict) –

  Other hyper-parameters of the Module, ref. corr. config file keyword arguments that have to be set are as follows:

  • groups: int, number of groups in the pointwise convolutional layer(s).

  • norm: bool or str or Module, normalization layer.

  • activation: str or Module, activation layer.

  • bias: bool, whether to use bias in the convolutional layer(s).

  • width_multiplier: float, multiplier of the number of output channels of the pointwise convolution.

  • stem: CFG, config of the stem block, with the following keys:

    • num_filters: int or Sequence[int], number of filters in the first convolutional layer(s).

    • filter_lengths: int or Sequence[int], filter lengths (kernel sizes) in the first convolutional layer(s).

    • subsample_lengths: int or Sequence[int], subsample lengths (strides) in the first convolutional layer(s).

  • inv_res: CFG, Config of the inverted residual blocks, with the following keys:

    • expansions: Sequence[int], expansion ratios of the inverted residual blocks.

    • out_channels: Sequence[int], number of output channels in each block.

    • n_blocks: Sequence[int], number of inverted residual blocks.

    • strides: Sequence[int], strides of the inverted residual blocks.

    • filter_lengths: Sequence[int], filter lengths (kernel sizes) in each block.

  • exit_flow: CFG, Config of the exit flow blocks, with the following keys:

    • num_filters: int or Sequence[int], number of filters in the final convolutional layer(s).

    • filter_lengths: int or Sequence[int], filter lengths (kernel sizes) in the final convolutional layer(s).

    • subsample_lengths: int or Sequence[int], subsample lengths (strides) in the final convolutional layer(s).

References

1

Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L. C. (2018). Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4510-4520).

2

https://github.com/pytorch/vision/blob/master/torchvision/models/mobilenetv2.py

3

https://github.com/keras-team/keras-applications/blob/master/keras_applications/mobilenet_v2.py

compute_output_shape(seq_len: Optional[int] = None, batch_size: Optional[int] = None) → Sequence[Optional[int]]

Compute the output shape of the model.

Parameters

• seq_len (int, optional) – Length of the input tensors.

• batch_size (int, optional) – Batch size of the input tensors.

Returns

output_shape – The output shape of the module.

Return type

sequence