Bayer

Last updated January 31, 2026

432

Innovation Areas

13,147

Inventors

104

Collaborations

Neural network weight compression architecture: BayerRecent Research Landscape

High computational overhead in deep neural networks limits real-time audio processing on edge devices. These innovations utilize folding structures and channelwise feature reorientation to reduce model complexity while maintaining signal robustness.

What technical problems is Bayer addressing in Neural network weight compression architecture?

Unreliable environmental perception accuracy

(29)evidences

Unreliable classification of road objects and poor quality of training labels lead to unsafe automated driving decisions. Improving perception accuracy ensures robust vehicle operation in complex real-world scenarios.

Excessive onboard computational demand

(4)evidences

High memory and processing requirements prevent deployment on resource-constrained hardware. Reducing these demands allows complex models to function in real-time environments.