CVMar 27, 2020

Dynamic Region-Aware Convolution

arXiv:2003.12243v3139 citations
Originality Incremental advance
AI Analysis

This is an incremental improvement for efficient computer vision networks, enhancing representation ability while maintaining computational cost.

The paper tackles the problem of modeling semantic variations in convolutional neural networks by proposing Dynamic Region-Aware Convolution (DRConv), which assigns multiple filters to spatial regions with similar features, resulting in a 6.3% relative improvement on ImageNet classification with ShuffleNetV2-0.5x achieving 67.1% accuracy at 46M multiply-adds.

We propose a new convolution called Dynamic Region-Aware Convolution (DRConv), which can automatically assign multiple filters to corresponding spatial regions where features have similar representation. In this way, DRConv outperforms standard convolution in modeling semantic variations. Standard convolutional layer can increase the number of filers to extract more visual elements but results in high computational cost. More gracefully, our DRConv transfers the increasing channel-wise filters to spatial dimension with learnable instructor, which not only improve representation ability of convolution, but also maintains computational cost and the translation-invariance as standard convolution dose. DRConv is an effective and elegant method for handling complex and variable spatial information distribution. It can substitute standard convolution in any existing networks for its plug-and-play property, especially to power convolution layers in efficient networks. We evaluate DRConv on a wide range of models (MobileNet series, ShuffleNetV2, etc.) and tasks (Classification, Face Recognition, Detection and Segmentation). On ImageNet classification, DRConv-based ShuffleNetV2-0.5x achieves state-of-the-art performance of 67.1% at 46M multiply-adds level with 6.3% relative improvement.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes