Returns the calculated values of the losses for optimization. outsize_batch: if None it doen't apply the enhance loss

Returns true if the two lists (valid and training) have a common element.

A function to create all the grids of images to log in wandb. images: A dictionary of images

Returns the needed losses for evaluating our model

Returns the optim and its scheduler. opt: a dictionary of the yaml config file with the train key

Util function to add the file path of all the images to the list of names of the selected images that will form the valid ones.

Given an image, it returns a center cropped version of size [cropx,cropy]

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

x_in: [b,h,w,c] # input_feature illu_fea: [b,h,w,c] # mask shift? 为什么是 b, h, w, c? return out: [b,h,w,c]

x: [b,h,w,c] return out: [b,h,w,c]

x: [b,c,h,w] illu_fea: [b,c,h,w] return out: [b,c,h,w]

x: [b,c,h,w] x是feature, 不是image illu_fea: [b,c,h,w] return out: [b,c,h,w]

x: [b,c,h,w] return out:[b,c,h,w]

Args: pred (Tensor): of shape (N, C, H, W). Predicted tensor. target (Tensor): of shape (N, C, H, W). Ground truth te

Args: pred (Tensor): of shape (N, C, H, W). Predicted tensor. target (Tensor): of shape (N, C, H, W). Ground truth te

Args: pred (Tensor): of shape (N, C, H, W). Predicted tensor. target (Tensor): of shape (N, C, H, W). Ground truth te

Args: pred (Tensor): of shape (N, C, H, W). Predicted tensor. target (Tensor): of shape (N, C, H, W). Ground truth te

Args: pred (Tensor): of shape (N, C, H, W). Predicted tensor. target (Tensor): of shape (N, C, H, W). Ground truth te

Inferences the video frames and constructs a new video. The result video is a composition of the original and the process ones.

Initiates wandb if needed. opt: a dictionary from the yaml config

Creates a logging dict to log results in wandb.

This function makes the same random sort to each list, so that they are sorted and the pairs are maintained.

Returns the loaded weights of model and optimizer if resume flag is True

Save the .pt of the model after each epoch.

Save tensor as PIL image.

A function that shuffles all the Distributed samplers in the loaders.

Create a weighted version of a given loss function. To use this decorator, the loss function must have the signature like `loss_func(pred, ta