human pose estimation / keypoint detection
AlphaPose is a real-time multi-person 2D full-body pose estimation and tracking system supporting body, face, hand, and foot keypoints.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Bottom-up multi-person pose estimation; different paradigm from top-down, good for crowded scenes without a separate detector.
This is an official implementation of our CVPR 2020 paper "HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation" (https://arxiv.org/abs/1908.10357)
CPU-optimized OpenPose-style bottom-up pose estimation; choose it when GPU is unavailable and speed matters.
Fast and accurate human pose estimation in PyTorch. Contains implementation of "Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose" paper.
The original bottom-up real-time keypoint library; prefer it for C++ integration and body+hand+face detection with a different architecture.
OpenPose: Real-time multi-person keypoint detection library for body, face, hands, and foot estimation
Center-based detection and pose estimation; a different approach useful for multi-task models on embedded devices.
Object detection, 3D detection, and pose estimation using center point detection:
TensorRT-accelerated real-time pose estimation; ideal for Jetson/NVIDIA edge devices with no tracking requirement.
Real-time pose estimation accelerated with NVIDIA TensorRT
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
A top-down single-person pose network (HRNet); choose it if you only need a lightweight backbone, not tracking.
The project is an official implementation of our CVPR2019 paper "Deep High-Resolution Representation Learning for Human Pose Estimation"
Simple Baselines top-down pose estimation; legacy code, useful only as a simple baseline without multi-person pipeline.
The project is an official implement of our ECCV2018 paper "Simple Baselines for Human Pose Estimation and Tracking(https://arxiv.org/abs/1804.06208)"
Whole-body pose estimation with distillation; current accuracy but no tracking, for users who need only keypoint inference.
"Effective Whole-body Pose Estimation with Two-stages Distillation" (ICCV 2023, CV4Metaverse Workshop)
Vision Transformer-based pose estimation; strong accuracy but no tracking, good for replacing the pose network in your own pipeline.
The official repo for [NeurIPS'22] "ViTPose: Simple Vision Transformer Baselines for Human Pose Estimation" and [TPAMI'23] "ViTPose++: Vision Transformer for Generic Body Pose Estimation"
A general pose estimation library, but stale; potentially useful for custom training pipelines in TensorLayer.
Library for Fast and Flexible Human Pose Estimation
Old TensorFlow top-down multi-person pose estimator; only relevant for reproducing CPN results.
[CVPR 2018] Cascaded Pyramid Network for Multi-Person Pose Estimation
PyTorch reimplementation of OpenPose; outdated but a simple bottom-up baseline.
The original Caffe OpenPose code; legacy version, only for historical compatibility.
Code repo for realtime multi-person pose estimation in CVPR'17 (Oral)
A PyTorch toolkit for 2D pose estimation; older and less featureful than AlphaPose.
A PyTorch toolkit for 2D Human Pose Estimation.
TensorFlow implementation of convolutional pose machines; a historical single-person method.
Very old TensorFlow pose estimation; only for legacy research.
Human Pose estimation with TensorFlow framework
These projects were analysed and named AlphaPose among their alternatives. The relationship is not symmetric — how AlphaPose rates them is a separate judgement, made when AlphaPose is analysed in its own right.
calls AlphaPose “Top-down multi-person pose and tracking with better accuracy on standard benchmarks, if you prefer top-down over OpenPose's speed.”
OpenPose: Real-time multi-person keypoint detection library for body, face, hands, and foot estimation
calls AlphaPose “Choose when you want a complete multi-person 2D pose-and-tracking system with an integrated detector instead of CenterNet's center-based component.”
Object detection, 3D detection, and pose estimation using center point detection:
calls AlphaPose “choose for accurate multi-person pose estimation, but it is stale and narrower than the seed.”
Ultralytics YOLO26, YOLO11, YOLOv8 — object detection, instance segmentation, semantic segmentation, image classification, pose estimation, object tracking