espnet.utils.spec_augment.dense_image_warp

source

espnet.utils.spec_augment.dense_image_warp(image, flow)

Image warping using per-pixel flow vectors.

Apply a non-linear warp to the image, where the warp is specified by a dense flow field of offset vectors that define the correspondences of pixel values in the output image back to locations in the source image. Specifically, the pixel value at output[b, j, i, c] is images[b, j - flow[b, j, i, 0], i - flow[b, j, i, 1], c]. The locations specified by this formula do not necessarily map to an int index. Therefore, the pixel value is obtained by bilinear interpolation of the 4 nearest pixels around (b, j - flow[b, j, i, 0], i - flow[b, j, i, 1]). For locations outside of the image, we use the nearest pixel values at the image boundary. Args: image: 4-D float Tensor with shape [batch, height, width, channels]. flow: A 4-D float Tensor with shape [batch, height, width, 2]. name: A name for the operation (optional). Note that image and flow can be of type tf.half, tf.float32, or tf.float64, and do not necessarily have to be the same type. Returns: A 4-D float Tensor with shape`[batch, height, width, channels]` and same type as input image. Raises: ValueError: if height < 2 or width < 2 or the inputs have the wrong number of dimensions.