USM sharpening. borrowed from real-ESRGAN Input image: I; Blurry image: B. 1. K = I + weight * (I - B) 2. Mask = 1 if abs(I - B) > thresho

generate an anisotropic Gaussian kernel Args: ksize : e.g., 15, kernel size theta : [0, pi], rotation angle range l1

generate an anisotropic Gaussian kernel Args: ksize : e.g., 15, kernel size theta : [0, pi], rotation angle range l1

Rezise the sample to ensure the given size. Keeps aspect ratio. Args: sample (dict): sample size (tuple): image size Returns

Kai Zhang (github: https://github.com/cszn)

bgr version of rgb2ycbcr only_y: only return Y channel Input: uint8, [0, 255] float, [0, 1]

Evaluate a function without caching intermediate activations, allowing for reduced memory at the expense of extra compute in the backward pas

Compute metrics for 'pred' compared to 'gt' Args: gt (numpy.ndarray): Ground truth values pred (numpy.ndarray): Predicted values

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

Reverses the imagenet normalization applied to the input. Args: x (torch.Tensor - shape(N,3,H,W)): input tensor Returns: tor

r"""Downsample2D a batch of 2D images with the given filter. Accepts a batch of 2D images of the shape `[N, C, H, W]` or `[N, H, W, C]` and downsa

boxes should be the output from dataset, which is a batch of bounding boxes

Forward pass. Args: x (tensor): input data (image) Returns: tensor: depth

Main entry point to get the config for the model. Args: model_name (str): name of the desired model. mode (str, optional): "train

Intrinsics for a pinhole camera model. Assume fov of 55 degrees and central principal point.

Learning rate configuration for different layers of the model Args: lr (float) : Base learning rate Returns:

Find and parse the .json config file for the model. Args: model_name (str): name of the model. The config file should be named config_{mo

This matches the implementation in Denoising Diffusion Probabilistic Models: From Fairseq. Build sinusoidal embeddings. This matches

imgs: list, N images of size WxHxC