姿态估计调研

综述

1. 人体姿态估计的过去，现在，未来
2. 重新思考人体姿态估计 Rethinking Human Pose Estimation
3. GitHub：人体姿态估计最全资料集锦
4. awesome-human-pose-estimation

有代码的论文

2D pose estimation

• Deep pose

• open pose

RGBD

• 3D Human Pose Estimation in RGBD Images for Robotic Task Learning

3D Human Pose

• Lifting from the Deep: Convolutional 3D Pose Estimation from a Single Image
• 3D human pose estimation in video with temporal convolutions and semi-supervised training#SOTA

Multi-view 3D Human Pose

• Self-Supervised Learning of 3D Human Pose using Multi-view Geometry#13 best
• Learnable Triangulation of Human Pose#SOTA
• Cross View Fusion for 3D Human Pose Estimation#SOTA
• Unsupervised Geometry-Aware Representation for 3D Human Pose Estimation

最相关的论文

• A Multi-view RGB-D Approach for Human Pose Estimation in Operating Rooms
• MVOR: A Multi-view RGB-D Operating Room Dataset for 2D and 3D Human Pose Estimation

数据集

影像精度的因素

• 环境
• 着装

Advantage of RGBD multi-view 3d pose estimation

• enables us to back-project points that are only visible in one view, while in multi-view RGB systems, points should be visible in at least two views.
• reduce the risk of occlusions
• In a multi-view RGB system, correspondences across views are traditionally established by relying on appearance similarity and triangulation [6, 14, 2, 7], which is unreliable in OR environments containing many surfaces that are visually similar