Learnable image transformation training methods and systems in graphics rendering

A method of training a shader for a frame transformation pipeline which is configured to transform rendered frames to produce enhanced frames comprising visual characteristics exhibited in a set of target images comprises: receiving input images and target images s 200; applying each shader to the input images to obtain candidate frames, and calculating, at a parametrized discriminator, a similarity indication between characteristics of the candidate frames and the target images S204. The method further comprises, in dependence on the indication, a parameter update step s206 to parameters of the discriminator and to one or more parametrized mathematical functions, wherein the parameter update step is configured to derive parameters for the parametrized mathematical function so that the shaders are arranged to impose their respective visual characteristics in dependence on an extent to which visual characteristic is exhibited in the target images. The indication is an objective loss value calculated using an adversarial loss function having two components indicative of accuracies of the discriminator and pipeline and the update step is performed by a generative adversarial network (GAN). The shaders can produce lens blur, colour mapping, distortion, bloom, sensor noise, and film grain.