Integral Human Pose Regression

By Xiao Sun, Bin Xiao, Shuang Liang, Yichen Wei. 2017

942 citations recorded by the source corpus.

State-of-the-art human pose estimation methods are based on heat map representation. In spite of the good performance, the representation has a few issues in nature, such as not differentiable and quantization error. This work shows that a simple integral operation relates and unifies the heat map representation and joint regression, thus avoiding the above issues. It is differentiable, efficient, and compatible with any heat map based methods. Its effectiveness is convincingly validated via comprehensive ablation experiments under various settings, specifically on 3D pose estimation, for the first time.

Subfields: Computer Vision and Pattern Recognition, 3D Vision & Pose, Human-Computer Interaction

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In 3D Vision & Pose

In Human-Computer Interaction

Citations in the corpus

Objects as Points (2019, neutral): “This is independently observed in keypoint regression [49,50].”
Deep High-Resolution Representation Learning for Visual Recognition (2019, neutral): “7 Integral Pose Regression [132] ResNet-101 256× 256 45.”
kPAM: KeyPoint Affordances for Category-Level Robotic Manipulation (2019, methodology): “The 2-D image coordinates and depth value are extracted using the integral operation [21].”
Deep High-Resolution Representation Learning for Human Pose Estimation (2019, neutral): “Integral Pose Regression [58] ResNet-101 256× 256 45.”
TDAN: Temporally-Deformable Alignment Network for Video Super-Resolution (2018, methodology): “It has been applied to address several high-level vision tasks such as object detection [4], video object detection [2], semantic seg-𝐼 mentation [4], and human pose estimation [24].”
HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation (2019, methodology): “Top-down methods [34, 9, 16, 42, 38, 40, 39, 16] take a dependency on person detector to detect person instances each with a bounding box and then reduce the problem to a sim-Heatmap [33, 30].”
Semantic Graph Convolutional Networks for 3D Human Pose Regression (2019, methodology): “As described in [54], the integral loss is used.”
Self-Supervised Learning of 3D Human Pose Using Multi-View Geometry (2019, methodology): “These analyses prove the stability of PSS. Implementation details We use the Integral Pose [36] architecture for both 2D and 3D branches with a ResNet-50 [13] backend.”
Monocular Total Capture: Posing Face, Body, and Hands in the Wild (2018, methodology): “Direct estimation methods predict 3D human pose directly from images, in the form of direct coordinate regression [46, 55, 56], voxel [42, 32, 66] or depth map [73].”
Learning to Reconstruct 3D Human Pose and Shape via Model-Fitting in the Loop (2019, background): “Recent works have made significant advances in the frontier of skeleton-based 3D human pose estimation from single images, with many approaches achieving impressive results [21, 23, 29, 33, 35, 45].”
MixSTE: Seq2seq Mixed Spatio-Temporal Encoder for 3D Human Pose Estimation in Video (2022, methodology): “Some methods [31, 37, 39] followed this manner but required a high computation cost due to regressing directly from the image space.”
I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image (2020, methodology): “To address the limitations, recent state-of-the-art 3D human pose estimation methods [32, 33, 44], which localize 3D human joint coordinates without mesh vertex coordinates, utilize the heatmap as the target representation of their networks.”
Sim2real transfer learning for 3D human pose estimation: motion to the rescue (2019, background): “Other works regress poses directly from pixels [46, 55, 56, 68, 71, 77, 78, 96], which generally relies on having a good match between training and testing.”
AlphaPose: Whole-Body Regional Multi-Person Pose Estimation and Tracking in Real-Time (2022, methodology): “Prior works [45], [46] adopt soft-max operation, which works well in single-person pose estimation but remains a large performance gap with the state-of-the-arts in multiperson pose estimation [31], [45], [60].”
Learnable Triangulation of Human Pose (2019, methodology): “6M [3], namely [16].”
Pose Recognition with Cascade Transformers (2021, methodology): “The field of human pose regression has witnessed a continuing progress [1, 32, 36, 22, 17, 2, 25, 29, 24, 6, 5, 28, 41, 23], in particular with the advancing of the deep learning technologies [18, 12, 14].”
Human Pose Regression with Residual Log-likelihood Estimation (2021, methodology): “It is seen that RLE reduces the error of single-stage regression baseline by 1.5 mm and the heatmap-based Integral Pose [54] by 0.6 mm.”