Image quantization using machine learning

A system for performing image quantization via machine learning, the system comprising:

• an input interface configured to receive input images from a sensor;
• and
• one or more processors in communication with the input interface, the one or more processors programmed to:
• (i) utilize an image-to-image machine-learning model to assign each pixel of the input images a new pixel color,
• (ii) utilize a mixer model to map each new pixel color to one of a fixed number of colors to produce a quantized image corresponding to each input image,
• (iii) align each input image with its corresponding quantized image via a pre-trained reference model to determine a loss function,
• (iv) backpropogate the loss function into the image-to-image machine learning model and the mixer model, and
• (v) output a trained image-to-image machine learning model and a trained mixer model after repeating (i)-(iv) until convergence.

A vehicle configured to utilize an image quantization machine-learning model, the vehicle comprising:

• a sensor configured to capture an image;
• and
• one or more processors communicatively coupled to the sensor and to memory, wherein the one or more processors are configured to utilize an image quantization machine-learning model to process the image, wherein the image quantization machine-learning model is trained according to the following steps:
• receiving a plurality of input images, wherein each input image includes a plurality of pixels;
• assigning each pixel a new pixel color via an image-to-image machine-learning model;
• converting each new pixel color to one of a fixed number of colors via a mixer model to produce a plurality of quantized images, each quantized image corresponding to one of the input images;
• computing a loss function associated with an alignment of each input image with its corresponding quantized image via a pre-trained reference model;
• updating one or more parameters of the image-to-image machine-learning model and the mixer model based on the loss function;
• and
• outputting a trained image quantization machine-learning system utilizing the updated parameters and upon convergence.