Scaling Wide Residual Networks for Panoptic Segmentation

TL;DR

This paper revisits and scales Wide Residual Networks for panoptic segmentation, demonstrating that simple architectural adjustments and scaling strategies significantly improve state-of-the-art results across various model regimes.

Contribution

It introduces SWideRNets, a family of scaled Wide-ResNets optimized for panoptic segmentation, with a straightforward scaling scheme that enhances performance.

Findings

Significant performance improvements on panoptic segmentation datasets.

Effective use of simple architectural modifications like SE and atrous convolutions.

Successful identification of high-performing models through grid search.

Abstract

The Wide Residual Networks (Wide-ResNets), a shallow but wide model variant of the Residual Networks (ResNets) by stacking a small number of residual blocks with large channel sizes, have demonstrated outstanding performance on multiple dense prediction tasks. However, since proposed, the Wide-ResNet architecture has barely evolved over the years. In this work, we revisit its architecture design for the recent challenging panoptic segmentation task, which aims to unify semantic segmentation and instance segmentation. A baseline model is obtained by incorporating the simple and effective Squeeze-and-Excitation and Switchable Atrous Convolution to the Wide-ResNets. Its network capacity is further scaled up or down by adjusting the width (i.e., channel size) and depth (i.e., number of layers), resulting in a family of SWideRNets (short for Scaling Wide Residual Networks). We demonstrate that such a simple scaling scheme, coupled with grid search, identifies several SWideRNets that significantly advance state-of-the-art performance on panoptic segmentation datasets in both the fast model regime and strong model regime.