Allows the creation of lists of layers, which is not supported by PyTorch in general. Detailed discussion here: https://discuss.pytor

Create a single hourglass

PyTorch U-Net model using ResNet(34, 101 or 152) encoder. UNet: https://arxiv.org/abs/1505.04597 ResNet: https://arxiv.org/abs/1512.03385

Create the stacked hourglass network

Tool Dataset constructed using the PyTorch built-in functionalities

Residual block without bottleneck

Residual block with bottleneck, suggested by Kaiming He et al., CVPR 2016

Residual block with bottleneck, improvement suggested by Kaiming He et al., ECCV 2016

Simple Convolutional Block consisting of conv - bn - ReLU layers

r""" FReLU formulation. The funnel condition has a window size of kxk. (k=3 by default)

Create the stacked hourglass network

Horizontally flip the given image and ground truth randomly with a probability of 0.5.

Scale (zoom-in, zoom-out) and Rotate the image and the ground truth. Args: maxRot (float): maximum rotation angle to be added maxS

Convert ndarrays in sample to Tensors.

PyTorch U-Net model using ResNet(34, 101 or 152) encoder. UNet: https://arxiv.org/abs/1505.04597 ResNet: https://arxiv.org/abs/1512.03385