sleap.nn.architectures.hrnet

(Higher)HRNet backbone.

This implementation is based on the PyTorch implementation of HRNet, modified to implement HigherHRNet’s configuration and new deconvolution heads.

Refs: https://arxiv.org/pdf/1902.09212.pdf https://arxiv.org/pdf/1908.10357.pdf

class sleap.nn.architectures.hrnet.HigherHRNet(C: int = 18, initial_downsampling_steps: int = 1, n_deconv_modules: int = 1, bottleneck: bool = False, deconv_filters: int = 256, bilinear_upsampling: bool = False, stem_filters: int = 64)

HigherHRNet backbone.

C

The variant of HRNet to use. The most common is HRNet32, which has ~30M params. This number is effectively the number of filters at the highest resolution output.

Type:

int

initial_downsampling_steps

Number of initial downsampling steps at the stem. Decrease if this introduces too much loss of resolution from the initial images.

Type:

int

n_deconv_modules

Number of upsampling steps to perform at the head. If this is equal to initial_downsampling_steps, the output will be at the same scale as the input.

Type:

int

bottleneck

If True, uses bottleneck blocks instead of simple residual blocks.

Type:

bool

deconv_filters

Number of filters to use in deconv blocks if using transposed convolutions.

Type:

int

bilinear_upsampling

Use bilinear upsampling instead of transposed convolutions at the output heads.

Type:

bool

property down_blocks

Returns the number of downsampling steps in the model.

output(x_in, n_output_channels)

Builds the layers for this backbone and return the output tensor.

Parameters:

• x_in – Input 4D tf.Tensor.

• n_output_channels – The number of final output channels.

Returns:

A tf.keras.model whose outputs are a list of tf.Tensors

at each scale of the deconv_modules.

Return type:

higher_hrnet_model

property output_scale

Returns relative scaling factor of this backbone.

sleap.nn.architectures.hrnet.adjust_prefix(name_prefix)

Adds a delimiter if the prefix is not empty.

sleap.nn.architectures.hrnet.bottleneck_block(x_in, filters, expansion_rate=4, name_prefix=None)

Creates a convolutional block with bottleneck.

sleap.nn.architectures.hrnet.simple_block(x_in, filters, stride=1, downsampling_layer=None, name_prefix=None)

Creates a basic residual convolutional block.