Pyramid Attention Network for Semantic Segmentation

Abstract

A Pyramid Attention Network(PAN) is proposed to exploit the impact of global contextual information in semantic segmentation. Different from most existing works, we combine attention mechanism and spatial pyramid to extract precise dense features for pixel labeling instead of complicated dilated convolution and artificially designed decoder networks. Specifically, we introduce a Feature Pyramid Attention module to perform spatial pyramid attention structure on high-level output and combining global pooling to learn a better feature representation, and a Global Attention Upsample module on each decoder layer to provide global context as a guidance of low-level features to select category localization details. The proposed approach achieves state-of-the-art performance on PASCAL VOC 2012 and Cityscapes benchmarks with a new record of mIoU accuracy 84.0% on PASCAL VOC 2012, while training without COCO dataset.

References in corpus (13)

• Rethinking Atrous Convolution for Semantic Image Segmentation
• ParseNet: Looking Wider to See Better
• The Cityscapes Dataset for Semantic Urban Scene Understanding
• Learning Deconvolution Network for Semantic Segmentation
• Stacked Hourglass Networks for Human Pose Estimation
• Path Aggregation Network for Instance Segmentation
• Beyond Skip Connections: Top-Down Modulation for Object Detection
• Residual Dense Network for Image Super-Resolution
• Semantic Image Segmentation via Deep Parsing Network
• Context Encoding for Semantic Segmentation
• Learning a Discriminative Feature Network for Semantic Segmentation
• Large Kernel Matters -- Improve Semantic Segmentation by Global Convolutional Network
• Automatic Real-time Background Cut for Portrait Videos

Cited by in corpus (45)

• SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers
• CHAOS Challenge -- Combined (CT-MR) Healthy Abdominal Organ Segmentation
• Improving Semantic Segmentation of Aerial Images Using Patch-based Attention
• Evolution of Image Segmentation using Deep Convolutional Neural Network: A Survey
• Hierarchical Multi-Scale Attention for Semantic Segmentation
• SCAttNet: Semantic Segmentation Network with Spatial and Channel Attention Mechanism for High-Resolution Remote Sensing Images
• Latent Correlation Representation Learning for Brain Tumor Segmentation with Missing MRI Modalities
• LoveDA: A Remote Sensing Land-Cover Dataset for Domain Adaptive Semantic Segmentation
• Robust Attentional Aggregation of Deep Feature Sets for Multi-view 3D Reconstruction
• DFANet: Deep Feature Aggregation for Real-Time Semantic Segmentation
• LeViT-UNet: Make Faster Encoders with Transformer for Medical Image Segmentation
• A global method to identify trees outside of closed-canopy forests with medium-resolution satellite imagery
• RAUNet: Residual Attention U-Net for Semantic Segmentation of Cataract Surgical Instruments
• Feature Pyramid Encoding Network for Real-time Semantic Segmentation
• Dual Refinement Feature Pyramid Networks for Object Detection
• Document Layout Analysis via Dynamic Residual Feature Fusion
• K-Hairstyle: A Large-scale Korean Hairstyle Dataset for Virtual Hair Editing and Hairstyle Classification
• RASNet: Segmentation for Tracking Surgical Instruments in Surgical Videos Using Refined Attention Segmentation Network
• DenseAttentionSeg: Segment Hands from Interacted Objects Using Depth Input
• A Survey on Deep Learning Methods for Semantic Image Segmentation in Real-Time
• Automatic Prostate Zonal Segmentation Using Fully Convolutional Network with Feature Pyramid Attention
• Semi-Supervised Segmentation of Salt Bodies in Seismic Images using an Ensemble of Convolutional Neural Networks
• DFPENet-geology: A Deep Learning Framework for High Precision Recognition and Segmentation of Co-seismic Landslides
• D2A U-Net: Automatic Segmentation of COVID-19 Lesions from CT Slices with Dilated Convolution and Dual Attention Mechanism
• Re-distributing Biased Pseudo Labels for Semi-supervised Semantic Segmentation: A Baseline Investigation
• CTNet: Context-based Tandem Network for Semantic Segmentation
• 3D Feature Pyramid Attention Module for Robust Visual Speech Recognition
• Learning to Segment Human Body Parts with Synthetically Trained Deep Convolutional Networks
• SINet: Extreme Lightweight Portrait Segmentation Networks with Spatial Squeeze Modules and Information Blocking Decoder
• Linear Context Transform Block
• Boundary Knowledge Translation based Reference Semantic Segmentation
• High resolution weakly supervised localization architectures for medical images
• Feature Fusion Use Unsupervised Prior Knowledge to Let Small Object Represent
• Real-Time Semantic Segmentation via Auto Depth, Downsampling Joint Decision and Feature Aggregation
• Exposing Semantic Segmentation Failures via Maximum Discrepancy Competition
• Adaptive Context Network for Scene Parsing
• Dense Dual-Path Network for Real-time Semantic Segmentation
• 3D Medical Multi-modal Segmentation Network Guided by Multi-source Correlation Constraint
• Automatic Segmentation of Gross Target Volume of Nasopharynx Cancer using Ensemble of Multiscale Deep Neural Networks with Spatial Attention
• Deep learning approach to left ventricular non-compaction measurement
• GSTO: Gated Scale-Transfer Operation for Multi-Scale Feature Learning in Pixel Labeling
• BiCANet: Bi-directional Contextual Aggregating Network for Image Semantic Segmentation
• An Abstraction Model for Semantic Segmentation Algorithms
• Differentiating Features for Scene Segmentation Based on Dedicated Attention Mechanisms
• Attention-Guided Lightweight Network for Real-Time Segmentation of Robotic Surgical Instruments