NPU for generating feature map based on coefficients and method thereof

An apparatus including:

• a semiconductor substrate on which an electrically conductive pattern is formed;
• at least one first memory electrically connected to the semiconductor substrate and configurable to store one or more first coefficients;
• and
• at least one neural processing unit (NPU) electrically connected to the semiconductor substrate and configurable to access the at least one first memory, the NPU comprising
• a first circuitry for at least one internal memory configurable to cache at least one feature map of a plurality of feature maps, and
• a second circuitry configured to generate a plurality of feature maps including a first feature map and the second feature map,
• wherein the first feature map is generated based on the one or more first coefficients in a first kernel filter,
• wherein the first kernel filter is generated based on a difference between a first kernel and a second kernel, and
• wherein the first kernel includes one or more first weights.

A method for operating an artificial neural network (ANN) model, comprising:

• receiving a first image input;
• generating a first feature map based on the first image input;
• and
• generating a second feature map based on the first feature map,
• wherein the generation of the first feature map is further based on one or more first coefficients in a first kernel filter,
• wherein the first kernel filter is generated based on a difference between a first kernel and a second kernel, and
• wherein the first kernel includes one or more first weights.