Browse Teach Curate

Data-Driven Enhancement Techniques for Neural Inertial Regression Networks

A systematic benchmark of 13 data-driven techniques - spanning network architecture, data augmentation, and data preprocessing - for improving deep neural networks that regress motion quantities from inertial (IMU) data. Evaluated across six platforms and over 1079 minutes of data, it finds that rotation and noise-addition augmentation give the most consistent gains and proposes standardized benchmarking strategies for the field.

Optimus Prime Text

Enhancement of Neural Inertial Regression Networks: A Data-Driven Perspective Victoria Khalfin

Deep learning has advanced inertial-sensing performance, but the field lacks a common benchmark for fairly comparing techniques that improve neural inertial regression networks. This paper defines an…