Temporal gradients of higher order effects to guide temporal accumulation

A method comprising:

• via a graphics core configured to perform a supersampling anti-aliasing operation via a neural network model:
• receiving, at an input block of the neural network model, a set of data including previous frame data, current frame data, velocity data, and jitter offset data, wherein the previous frame data includes pixels associated with lighting effects;
• pre-processing the set of data to generate pre-processed data, wherein pre-processing the set of data includes to warp pixels associated with the lighting effects in the previous frame via motion vectors generated for the pixels;
• providing pre-processed data to a feature extraction network of the neural network model;
• processing the pre-processed data at the feature extraction network via one or more encoder stages and one or more decoder stages;
• outputting tensor data from the feature extraction network to the output block;
• and
• generating an output image via an output block of the neural network model, the output image a supersampled, and anti-aliased output image.

A data processing system comprising:

• a memory device configured to include instructions;
• and
• a graphics processor coupled with the memory device, the graphics processor including a graphics core configured to execute the instructions, wherein the instructions are to configure the graphics core to perform a supersampling anti-aliasing operation via a mixed precision convolutional neural network, the graphics core including circuitry configured to:
• receive, at an input block of a neural network model, a set of data including previous frame data, current frame data, velocity data, and jitter offset data, wherein the previous frame data includes pixels associated with lighting effects;
• pre-process the set of data to generate pre-processed data, wherein to pre-process the set of data includes to warp pixels associated with the lighting effects in the previous frame via motion vectors generated for the pixels;
• provide pre-processed data to a feature extraction network of the neural network model;
• process the pre-processed data at the feature extraction network via one or more encoder stages and one or more decoder stages;
• output tensor data from the feature extraction network to the output block;
• and
• generate an output image via an output block of the neural network model, the output image a supersampled and anti-aliased output image.