Low Bitrates

Low bitrate compression aims to represent data (audio, images, video) using minimal storage or transmission bandwidth while maintaining acceptable quality. Current research heavily utilizes generative models, including diffusion models, variational autoencoders (VAEs), and generative adversarial networks (GANs), often incorporating techniques like conditional flow matching and map-assisted guidance to improve reconstruction fidelity at extremely low bitrates. This field is crucial for applications requiring efficient data transmission in resource-constrained environments, such as edge devices and low-bandwidth networks, and is driving advancements in both model architecture and algorithmic efficiency.