Inference with slide/whole style. Args: img (Tensor): The input image of shape (N, 3, H, W). img_meta (dict): Image i

Inference with slide/whole style. Args: img (Tensor): The input image of shape (N, 3, H, W). img_meta (dict): Image i

x: [B, H, W, N, C] Returns: (B*num_windows, num_heads, window_size*window_size*N, head_dim]

Returns: bool: whether the arguments contain CfgNode and should be considered forwarded to from_config.

Use `from_config` to obtain explicit arguments. Returns: dict: arguments to be used for cls.__init__

Wrapper for grid_sample, uses pixel coordinates

Wrapper for grid_sample, uses pixel coordinates

Wrapper for grid_sample, uses pixel coordinates

Wrapper for grid_sample, uses pixel coordinates

Wrapper for grid_sample, uses pixel coordinates

Wrapper for grid_sample, uses pixel coordinates

1D bilinear sampling along width dimension only (for stereo applications) Much faster than grid_sample for stereo where y is constant Ar

compute smooth l1 loss :param pred: disparity prediction [N,H,W] :param inputs: input data :param invalid_mask: inva

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

3x3 convolution with padding

3x3 convolution with padding

fmap1: [B, C, H, W] warped_fmap2: [B, C, H, W, num_disp] Returns: local cost: [B*H*W, num_disp, G]

cross_entropy. The wrapper function for :func:`F.cross_entropy` Args: pred (torch.Tensor): The prediction with shape (N, 1). labe

Dampen the learning rates. Arguments: step (int): The index of current step. (Default: None)

Prediction each pixel.

Prediction each pixel.

Transfer disparity map to color map Args: disp (numpy.array): disparity map in (Height, Width) layout, value range [0, 255] m