Pose Estimation Models logopeterindia.net
Knowledge Hub

Pose Estimation Models for Human Keypoint Detection

A directory of open-source models and libraries for detecting and analysing human body keypoints in images and video.

19 topics•5 categories•Updated October 2026

Real-time & Lightweight Models

4 entries

Fast models for edge devices and live video.

01MoveNetUltra-fast, accurate pose detection model from TensorFlow Hub. 02MediaPipe Pose LandmarkerGoogle's on-device pose landmark detection for mobile, web and desktop. 03Lightweight OpenPoseCompact PyTorch implementation of OpenPose for real-time use on CPUs. 04Ultralytics YOLO PoseYOLO models with keypoint heads for pose estimation.

Multi-person & Bottom-up Models

3 entries

Detect many people in one image.

05OpenPoseCMU's real-time multi-person system for body, hand, face and foot keypoints. 06HigherHRNetBottom-up high-resolution network for multi-person pose estimation. 07AlphaPoseAccurate multi-person pose estimation and tracking system.

Transformer & Video Models

3 entries

Attention-based and temporal approaches.

08TransPoseTransformer-based keypoint detection with interpretable attention. 09ViTPosePlain Vision Transformer baselines for pose estimation. 10DCPoseVideo pose estimation using dual-context information across frames.

Dense & Toolbox Projects

2 entries

Dense body surface mapping and toolkits.

11DensePoseMaps image pixels to a 3D body surface; the facebookresearch repository is archived. 12MMPoseOpenMMLab toolbox with many pose models, including RTMPose.

Related Pages

7 entries

Neighbouring directories.

13Computer VisionComputer vision hub. 14Deep LearningDeep learning frameworks and datasets. 15Machine LearningMachine learning hub. 16Data Sets for ML/DLDatasets for training and benchmarking. 17Image Data Analytics SolutionsImage analytics solutions. 18Video Data Analytics SolutionsVideo analytics solutions. 19Generative AI Platforms and ToolsGenerative AI companies and tools.

Frequently Asked Questions

What is pose estimation?

Pose estimation locates the keypoints of a body, such as shoulders, elbows and knees, in images or video so software can analyse posture and movement.

What is the difference between top-down and bottom-up methods?

Top-down methods detect each person first and then estimate keypoints; bottom-up methods find all keypoints first and group them into people, which scales better in crowds.

Which model suits real-time use on a phone?

MoveNet and MediaPipe Pose are designed for fast on-device inference; lightweight OpenPose and YOLO pose are alternatives.

What is DensePose?

A task and model that maps pixels of a person to a 3D body surface rather than a few keypoints; its Facebook Research repository is now archived.

Where are the applications?

Fitness and sports analysis, healthcare and rehabilitation, animation, retail analytics and human-computer interaction.