compute px error :param pred: disparity prediction [N,H,W] :param disp: ground truth disparity [N,H,W] :param loss_d

Produce N identical layers.

compute binary entropy loss on occlusion mask :param occ_pred: occlusion prediction, [N,H,W] :param inputs: input data

compute EPE :param pred: disparity prediction [N,H,W] :param disp: ground truth disparity [N,H,W] :param loss_dict:

compute IOU on occlusion :param pred: occlusion prediction [N,H,W] :param occ_mask: ground truth occlusion mask [N,H,W]

Compute occluded region on the left image border :param w: image width :param disp: left disparity :return: occ mask

Compute occluded region on the right image border :param w: image width :param disp: right disparity :return: occ mas

计算 monocular depth 和 ground truth depth 之间的 scale 和 shift. 参数: monocular_depth (torch.Tensor): 单目深度图，形状为 (H, W) 或 (N, H, W) gt_d

@disp_low: (b,1,h,w) 1/4 resolution @up_weights: (b,9,4*h,4*w) Image resolution

Calculate dice loss, which is proposed in `V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation <https://arxiv

Based on color histogram, convert the gray disp into color disp map. The histogram consists of 7 bins, value of each is e.g. [114.0, 185.0, 1

Warping by disparity Args: img: [B, 3, H, W] disp: [B, 1, H, W], positive padding_mode: 'zeros' or 'border' Returns:

Warping by disparity Args: img: [B, 3, H, W] disp: [B, 1, H, W], positive padding_mode: 'zeros' or 'border' Returns:

Returns: str: a yaml string representation of the config

Implements the "standard" process for log regression evaluation: The value of C is chosen by training on train_dataset and evaluating on

x_list contains a list of tensors to nest together and run

x_list contains a list of tensors to nest together and run