Resamples a uint8[height, width, channels] image tensor through a
grid_sample-style backward field, as produced by the UVDoc dewarping
model, into a new tensor of the same size.
grid is a float32 tensor of shape [2, gridHeight, gridWidth] (a
leading batch dimension of 1 is accepted): channel 0 holds x and channel 1
holds y, both normalized to [-1, 1] with align_corners = true. The grid
is bilinearly upsampled to the image size natively, so it may be much
coarser than the image. Sampling is cv::remap. Images of 1 to 4 channels
are supported. The caller owns every tensor involved.
Resamples a
uint8[height, width, channels]image tensor through agrid_sample-style backward field, as produced by the UVDoc dewarping model, into a new tensor of the same size.gridis afloat32tensor of shape[2, gridHeight, gridWidth](a leading batch dimension of 1 is accepted): channel 0 holds x and channel 1 holds y, both normalized to[-1, 1]withalign_corners = true. The grid is bilinearly upsampled to the image size natively, so it may be much coarser than the image. Sampling iscv::remap. Images of 1 to 4 channels are supported. The caller owns every tensor involved.