Create a 1D, 2D, or 3D average pooling module.

Args: x: HxWxC image, [0, 1] sf: down-scale factor Return: bicubicly downsampled LR image

Args: x: HxWxC image, [0, 1] sf: down-scale factor Return: bicubicly downsampled LR image

Converts a depth map to a color image. Args: value (torch.Tensor, numpy.ndarry): Input depth map. Shape: (H, W) or (1, H, W) or (1, 1, H,

Compute metrics of predicted depth maps. Applies cropping and masking as necessary or specified via arguments. Refer to compute_errors for more detail

Take an RGB image and depth map and place them side by side. This includes a proper normalization of the depth map for better visibility.

Crops the images according to the crop parameters. Args: images: RGB or depth images, shape (H, W, 3) or (H, W). crop_params: Cro

Crops the black border of the RGB. Args: rgb: RGB image, shape (H, W, 3). Returns: Crop parameters.

Returns parameters of all the ModuleDicts children that are exclusively used for the given bin configuration

Crops the white border of the RGB. Args: rgb: RGB image, shape (H, W, 3). Returns: Crop parameters.

Inference interface for the model Args: x (torch.Tensor): input tensor of shape (b, c, h, w) pad_input (bool,

Load state_dict into model, handling DataParallel and DistributedDataParallel. Also checks for "model" key in state_dict. DataParallel prefixes s

Loads weights to the model from a given resource. A resource can be of following types: 1. URL. Prefixed with "url::" e.g. url

Run the inference and interpolate. Args: device (torch.device): the torch device used model: the model used for inference

shift pixel for super-resolution with different scale factors Args: x: WxHxC or WxH sf: scale factor upper_left: shift dir

shift pixel for super-resolution with different scale factors Args: x: WxHxC or WxH sf: scale factor upper_left: shift dir

Create sinusoidal timestep embeddings. :param timesteps: a 1-D Tensor of N indices, one per batch element. These may be

Zero out the parameters of a module and return it.

Init. Args: width (int): desired output width height (int): desired output height resize_target (bool, opt

Compute log binomial distribution for n_classes Args: n_classes (int, optional): number of output classes. Defaults to 256.

Attractor layer for bin centers. Bin centers are bounded on the interval (min_depth, max_depth)

Spatial-Temporal Self-Attn -> Cross-Frame Gated-Attn -> Modulated Cross-Attn -> FF

Inference interface for the model Args: x (torch.Tensor): input tensor of shape (b, c, h, w) Returns: