Comparative Study of Parameter Selection for Enhanced Edge Inference for a Multi-Output Regression model for Head Pose Estimation

TL;DR

This paper investigates parameter selection and magnitude-based pruning to optimize a multi-output regression model for head pose estimation, achieving significant size reduction and improved accuracy for edge inference.

Contribution

It introduces an optimized pruning approach that enhances model efficiency and accuracy specifically for head pose estimation tasks.

Findings

Over 75% model size reduction achieved

Higher accuracy than original model

Effective parameter selection for edge inference

Abstract

Magnitude-based pruning is a technique used to optimise deep learning models for edge inference. We have achieved over 75% model size reduction with a higher accuracy than the original multi-output regression model for head-pose estimation.