Args: cfg_other (CfgNode): config to merge from.

Generate meshgrid in image scale Args: img: [B, _, H, W] homogeneous: whether to return homogeneous coordinates Return:

Generate meshgrid in image scale Args: img: [B, _, H, W] homogeneous: whether to return homogeneous coordinates Return:

Generate meshgrid in image scale Args: img: [B, _, H, W] homogeneous: whether to return homogeneous coordinates Return:

Calculate naive dice loss, the coefficient in the denominator is the first power instead of the second power. Args: pred (torch.Tenso

Normalize coordinates of image scale to [-1, 1] Args: grid: [B, 2, H, W]

Normalize coordinates of image scale to [-1, 1] Args: grid: [B, 2, H, W]

Normalize coordinates of image scale to [-1, 1] Args: grid: [B, 2, H, W]

Pairwise nearest neighbors for L2-normalized vectors. Uses Torch rather than Faiss to remain on GPU.

Args: inputs: the inputs to the model. It is a dict, which contains keys like "img1", "ïmg2", "disp".

Depth predictions have arbitrary scale - need to convert to a pixel disparity

Read .flo file in Middlebury format

"Obtain the mean of tensor on different GPUs.

Args: img: same as in __init__

cost_volume: [B*H*W,G,D], groupwise cost volume label_seed: [B*H*W,num_seed], integer disparity modals return: [B

Placeholder for single image test.

Simple test with single image.

Generate the valid flags of points of a single feature map. Args: featmap_size (tuple[int]): The size of feature maps, arrange as

Inference by sliding-window with overlap. If h_crop > h_img or w_crop > w_img, the small patch will be used to decode without padding

Inference by sliding-window with overlap. If h_crop > h_img or w_crop > w_img, the small patch will be used to decode without padding

Inference by sliding-window with overlap. If h_crop > h_img or w_crop > w_img, the small patch will be used to decode without padding

Tag the last checkpoint. Args: last_filename_basename (str): the basename of the last filename.

linearly sample source tensor along the last dimension input: source [N,D1,D2,D3...,Dn] sample_points [N,D1,D2,....,Dn-1,1]

Upsample flow field [H, W, 2] -> [2 * H/4, 2 * W, 2] using convex combination