check if all the image routes are correct

Creates a dictionary from the config yaml file.

Creates the model. opt: a dictionary from the yaml config key network

A function to create a grid of images to log in wandb.

opt: a dictionary from the yaml config key datasets

yaml orderedDict support

Apply the inference over each specific frame. If resize = True, resizes before inference.

Transform from numpy array [H,W,C] to torch tensor [B,C,H,W]

Creates a set of paths to save the model based on the config file.

This function runs over the multiple test loaders and returns the whole metrics.

Load the weights of the model.

Loads the values of the optimizer picking only the weights that are in the new model.

Loads the weights of a pretrained model, picking only the weights that are in the new model.

Normalize tensor to the range [0,1]

Pad the tensor to be multiple of some number (its size).

pad the tensor to be multiple of some number

Pad the tensor to be multiple of some number (its size).

Reduce loss as specified. Args: loss (Tensor): Elementwise loss tensor. reduction (str): Options are 'none', 'mean' and 'sum'.

Transform from torch tensor [B,C,H,W] to numpy array [H,W,C].

Apply element-wise weight and reduce loss. Args: loss (Tensor): Element-wise loss. weight (Tensor): Element-wise weights. Default

Given a (random) index. The dataloader selects the corresponding image path, and loads the image. Then it returns the image, after ap

- images_high: list of RGB images of normal-light used for training or testing the model - images_low: list of RGB images of low-ligh