Efficient Re-parameterization Operations Search for Easy-to-Deploy Network Based on Directional Evolutionary Strategy

TL;DR

This paper introduces an improved re-parameterization search space and an evolutionary NAS strategy to automatically discover better re-parameterization architectures, significantly enhancing CNN performance without relying on prior operation knowledge.

Contribution

It proposes a novel, expanded re-parameterization search space and a directional evolutionary strategy for NAS to optimize re-parameterization architectures automatically.

Findings

Improved ResNet-50 accuracy by 1.82% on ImageNet-1k.

Designed a flexible search space including more re-parameterization operations.

Achieved better results than baseline models under same training conditions.

Abstract

Structural re-parameterization (Rep) methods has achieved significant performance improvement on traditional convolutional network. Most current Rep methods rely on prior knowledge to select the reparameterization operations. However, the performance of architecture is limited by the type of operations and prior knowledge. To break this restriction, in this work, an improved re-parameterization search space is designed, which including more type of re-parameterization operations. Concretely, the performance of convolutional networks can be further improved by the search space. To effectively explore this search space, an automatic re-parameterization enhancement strategy is designed based on neural architecture search (NAS), which can search a excellent re-parameterization architecture. Besides, we visualize the output features of the architecture to analyze the reasons for the formation of the re-parameterization architecture. On public datasets, we achieve better results. Under the same training conditions as ResNet, we improve the accuracy of ResNet-50 by 1.82% on ImageNet-1k.