Conditional axial transformation layer for high fidelity image transformation

Devices and methods involve receiving an input image comprising an array of pixels, wherein the input image is associated with a first characteristic; a neural network is applied to transform an input image into an output image associated with a second characteristic, an encoded pixel is generated by an encoder for each pixel of an array of pixels of the input image, the array of encoded pixels is provided to a decoder, axial attention is applied by the decoder to decode a given pixel, the axial attention includes row attention or column attention applied to one or more previously decoded pixels in a row or column preceding the row or column associated with a given pixel, and wherein the row or column attention mixes information within the respective row or column and maintains independence between respective different rows or columns; and generating an output image through the neural network. 设备和方法涉及：接收包括像素阵列的输入图像，其中输入图像与第一特性相关联；应用神经网络以将输入图像变换为与第二特性相关联的输出图像，由编码器对输入图像的像素阵列的每个像素生成编码像素，向解码器提供编码像素阵列，由解码器应用轴向注意来解