Rethinking on Multi-Stage Networks for Human Pose Estimation

Abstract

Existing pose estimation approaches fall into two categories: single-stage and multi-stage methods. While multi-stage methods are seemingly more suited for the task, their performance in current practice is not as good as single-stage methods. This work studies this issue. We argue that the current multi-stage methods' unsatisfactory performance comes from the insufficiency in various design choices. We propose several improvements, including the single-stage module design, cross stage feature aggregation, and coarse-to-fine supervision. The resulting method establishes the new state-of-the-art on both MS COCO and MPII Human Pose dataset, justifying the effectiveness of a multi-stage architecture. The source code is publicly available for further research.

References in corpus (13)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• Associative Embedding: End-to-End Learning for Joint Detection and Grouping
• Human pose estimation via Convolutional Part Heatmap Regression
• DetNet: A Backbone network for Object Detection
• Human Pose Regression by Combining Indirect Part Detection and Contextual Information
• Simple Baselines for Human Pose Estimation and Tracking
• Multi-Context Attention for Human Pose Estimation
• Human Pose Estimation with Spatial Contextual Information
• Deep High-Resolution Representation Learning for Human Pose Estimation
• Self Adversarial Training for Human Pose Estimation
• Multi-Scale Structure-Aware Network for Human Pose Estimation

Cited by in corpus (38)

• A Survey of Deep Learning-based Object Detection
• Deep Learning-Based Human Pose Estimation: A Survey
• Key Points Estimation and Point Instance Segmentation Approach for Lane Detection
• TFPose: Direct Human Pose Estimation with Transformers
• Multi-Stage Progressive Image Restoration
• Self-supervised Keypoint Correspondences for Multi-Person Pose Estimation and Tracking in Videos
• Supervision by Registration and Triangulation for Landmark Detection
• Heatmap Regression via Randomized Rounding
• Simple and Lightweight Human Pose Estimation
• Learning Delicate Local Representations for Multi-Person Pose Estimation
• The Devil is in the Details: Delving into Unbiased Data Processing for Human Pose Estimation
• SPGNet: Semantic Prediction Guidance for Scene Parsing
• Cascade Feature Aggregation for Human Pose Estimation
• OmniPose: A Multi-Scale Framework for Multi-Person Pose Estimation
• EvoPose2D: Pushing the Boundaries of 2D Human Pose Estimation using Accelerated Neuroevolution with Weight Transfer
• LUVLi Face Alignment: Estimating Landmarks' Location, Uncertainty, and Visibility Likelihood
• Graph-PCNN: Two Stage Human Pose Estimation with Graph Pose Refinement
• AID: Pushing the Performance Boundary of Human Pose Estimation with Information Dropping Augmentation
• LightTrack: A Generic Framework for Online Top-Down Human Pose Tracking
• Simple Pose: Rethinking and Improving a Bottom-up Approach for Multi-Person Pose Estimation
• Multi-Stage HRNet: Multiple Stage High-Resolution Network for Human Pose Estimation
• Towards High Performance Human Keypoint Detection
• SMAP: Single-Shot Multi-Person Absolute 3D Pose Estimation
• TRB: A Novel Triplet Representation for Understanding 2D Human Body
• Aggregation and Finetuning for Clothes Landmark Detection
• Adversarial Semantic Data Augmentation for Human Pose Estimation
• Train Your Data Processor: Distribution-Aware and Error-Compensation Coordinate Decoding for Human Pose Estimation
• Adaptive Dilated Convolution For Human Pose Estimation
• PI-Net: Pose Interacting Network for Multi-Person Monocular 3D Pose Estimation
• Fast Uncertainty Quantification for Deep Object Pose Estimation
• Efficient Human Pose Estimation by Learning Deeply Aggregated Representations
• EfficientPose: Efficient Human Pose Estimation with Neural Architecture Search
• Orderly Dual-Teacher Knowledge Distillation for Lightweight Human Pose Estimation
• Regressive Domain Adaptation for Unsupervised Keypoint Detection
• Human Keypoint Detection by Progressive Context Refinement
• Towards Good Practices for Multi-Person Pose Estimation
• Self-Supervision and Spatial-Sequential Attention Based Loss for Multi-Person Pose Estimation
• Multi-Level Network for High-Speed Multi-Person Pose Estimation