Dynamic Dual Sampling Module for Fine-Grained Semantic Segmentation

TL;DR

This paper introduces a Dynamic Dual Sampling Module (DDSM) that enhances semantic segmentation by adaptively propagating semantic context to local details, improving boundary accuracy and overall performance.

Contribution

The novel DDSM method models dynamic affinity and propagates semantic context to local details, addressing the gap in exploring their interrelationship in segmentation.

Findings

Improves segmentation accuracy on Cityscapes and Camvid datasets.

Effectively preserves object boundaries in segmentation results.

Demonstrates efficiency and effectiveness of the proposed approach.

Abstract

Representation of semantic context and local details is the essential issue for building modern semantic segmentation models. However, the interrelationship between semantic context and local details is not well explored in previous works. In this paper, we propose a Dynamic Dual Sampling Module (DDSM) to conduct dynamic affinity modeling and propagate semantic context to local details, which yields a more discriminative representation. Specifically, a dynamic sampling strategy is used to sparsely sample representative pixels and channels in the higher layer, forming adaptive compact support for each pixel and channel in the lower layer. The sampled features with high semantics are aggregated according to the affinities and then propagated to detailed lower-layer features, leading to a fine-grained segmentation result with well-preserved boundaries. Experiment results on both Cityscapes and Camvid datasets validate the effectiveness and efficiency of the proposed approach. Code and models will be available at \url{x3https://github.com/Fantasticarl/DDSM}.