Optical Diffusion Models for Image Generation

TL;DR

This paper introduces an optical diffusion model that uses passive diffractive layers to perform image denoising, enabling fast, energy-efficient image generation by leveraging optical hardware instead of traditional neural networks.

Contribution

It presents a novel optical implementation of diffusion models using passive diffractive layers, reducing latency and energy consumption in image generation.

Findings

Optical diffusion model achieves high-speed image generation.

The method significantly reduces power consumption compared to electronic neural networks.

Passive optical layers effectively perform denoising without active computation.

Abstract

Diffusion models generate new samples by progressively decreasing the noise from the initially provided random distribution. This inference procedure generally utilizes a trained neural network numerous times to obtain the final output, creating significant latency and energy consumption on digital electronic hardware such as GPUs. In this study, we demonstrate that the propagation of a light beam through a semi-transparent medium can be programmed to implement a denoising diffusion model on image samples. This framework projects noisy image patterns through passive diffractive optical layers, which collectively only transmit the predicted noise term in the image. The optical transparent layers, which are trained with an online training approach, backpropagating the error to the analytical model of the system, are passive and kept the same across different steps of denoising. Hence this method enables high-speed image generation with minimal power consumption, benefiting from the bandwidth and energy efficiency of optical information processing.