System integrated machine-learning co-processing

A method of generating an augmented image comprising:

• receiving, at an image signal processor (ISP), raw color sensor data from a color sensor that is associated with a first type of camera;
• receiving, at the image signal processor (ISP), raw second sensor data from a second sensor that is associated with a second type of camera, wherein the first type of camera and second type of camera are different;
• generating a color image at the ISP using the raw color sensor data;
• generating a second image at the ISP using the second sensor data;
• communicating the color image and the second image from the ISP to a neural processing unit (NPU), wherein the NPU includes a machine-learning (ML) model trained to make an inference about an input image;
• receiving a ML inference about the color image from the NPU, wherein the NPU generated the ML inference using the second image as input;
• and
• communicating the ML inference from the ISP to a computing component.

A camera system comprising:

• a color image sensor;
• an image signal processor (ISP) communicatively coupled to the color image sensor;
• a neural processing unit (NPU) communicatively coupled to the ISP, wherein the NPU includes a machine-learning (ML) model trained to make an ML inference about a color image generated by the ISP;
• a steganography component programed to associate the ML inference with the color image using steganography to generate an augmented image;
• and
• an output port configured to output the augmented image comprising the color image and the ML inference.