Dual Attention Network for Scene Segmentation

Abstract

In this paper, we address the scene segmentation task by capturing rich contextual dependencies based on the selfattention mechanism. Unlike previous works that capture contexts by multi-scale features fusion, we propose a Dual Attention Networks (DANet) to adaptively integrate local features with their global dependencies. Specifically, we append two types of attention modules on top of traditional dilated FCN, which model the semantic interdependencies in spatial and channel dimensions respectively. The position attention module selectively aggregates the features at each position by a weighted sum of the features at all positions. Similar features would be related to each other regardless of their distances. Meanwhile, the channel attention module selectively emphasizes interdependent channel maps by integrating associated features among all channel maps. We sum the outputs of the two attention modules to further improve feature representation which contributes to more precise segmentation results. We achieve new state-of-the-art segmentation performance on three challenging scene segmentation datasets, i.e., Cityscapes, PASCAL Context and COCO Stuff dataset. In particular, a Mean IoU score of 81.5% on Cityscapes test set is achieved without using coarse data. We make the code and trained model publicly available at https://github.com/junfu1115/DANet

arXiv commentAccepted by CVPR2019

References in corpus (10)

• Rethinking Atrous Convolution for Semantic Image Segmentation
• Self-Attention Generative Adversarial Networks
• A Structured Self-attentive Sentence Embedding
• Stacked Hourglass Networks for Human Pose Estimation
• OCNet: Object Context Network for Scene Parsing
• Wider or Deeper: Revisiting the ResNet Model for Visual Recognition
• Semantic Image Segmentation via Deep Parsing Network
• Stacked Deconvolutional Network for Semantic Segmentation
• Relation Networks for Object Detection
• COCO-Stuff: Thing and Stuff Classes in Context

Cited by in corpus (37)

• Deep High-Resolution Representation Learning for Visual Recognition
• Dual Graph Convolutional Network for Semantic Segmentation
• An Empirical Study of Spatial Attention Mechanisms in Deep Networks
• Weak Supervision for Generating Pixel-Level Annotations in Scene Text Segmentation
• Comprehensive Attention Self-Distillation for Weakly-Supervised Object Detection
• DyNet: Dynamic Convolution for Accelerating Convolutional Neural Networks
• Attention-guided Context Feature Pyramid Network for Object Detection
• ABD-Net: Attentive but Diverse Person Re-Identification
• SLSNet: Skin lesion segmentation using a lightweight generative adversarial network
• Relation-Aware Global Attention for Person Re-identification
• EANet: Enhancing Alignment for Cross-Domain Person Re-identification
• GFF: Gated Fully Fusion for Semantic Segmentation
• An Efficient Solution for Breast Tumor Segmentation and Classification in Ultrasound Images Using Deep Adversarial Learning
• Attention-guided Unified Network for Panoptic Segmentation
• Spatial Pyramid Based Graph Reasoning for Semantic Segmentation
• Squeeze-and-Attention Networks for Semantic Segmentation
• A Novel Upsampling and Context Convolution for Image Semantic Segmentation
• Semantic Flow for Fast and Accurate Scene Parsing
• Example-Guided Style Consistent Image Synthesis from Semantic Labeling
• Tensor Low-Rank Reconstruction for Semantic Segmentation
• Transformer Meets Convolution: A Bilateral Awareness Network for Semantic Segmentation of Very Fine Resolution Urban Scene Images
• An Attention-Aided Deep Learning Framework for Massive MIMO Channel Estimation
• Deep Angular Embedding and Feature Correlation Attention for Breast MRI Cancer Analysis
• Improving Object Detection with Inverted Attention
• OGNet: Salient Object Detection with Output-guided Attention Module
• Local block-wise self attention for normal organ segmentation
• Consensus Feature Network for Scene Parsing
• Non-linear and Selective Fusion of Cross-Modal Images
• Two-Stream Video Classification with Cross-Modality Attention
• Single Image Depth Estimation Trained via Depth from Defocus Cues
• Detecting Reflections by Combining Semantic and Instance Segmentation
• High-Order Paired-ASPP Networks for Semantic Segmenation
• Fast and Accurate Scene Parsing via Bi-direction Alignment Networks
• Retrieval and Localization with Observation Constraints
• Dynamic Dual Sampling Module for Fine-Grained Semantic Segmentation
• Learning Parallax Attention for Stereo Image Super-Resolution
• Cross Attention Network for Semantic Segmentation