Human pose and shape estimation
Also written: Human body pose and shape estimation
Estimating a person's body pose and physical shape from images or video.
Human pose and shape estimation is about understanding a person's body from visual data. It involves figuring out where the main joints of the body are, like the shoulders, elbows, and knees, and also understanding the overall size and form of the person.
This technology can be used in many ways. For example, it could help a video game character mimic your movements, or it could be used in virtual try-on applications to see how clothes would fit your body shape.
Human pose and shape estimation refers to the process of inferring the 3D spatial configuration of human body joints (pose) and the body's geometric dimensions (shape) from visual input, typically images or video. This involves detecting keypoints like wrists, elbows, and hips, and often reconstructing a parametric body model (e.g., SMPL) to represent the full shape. Methods range from direct regression of joint coordinates or shape parameters to more complex approaches involving volumetric representations or graph neural networks.
Recent advancements in 3D reconstruction from single images, like Meta AI's SAM 3D, highlight the growing capability of AI to understand and represent human bodies in three dimensions. This progress is driven by more powerful foundation models and improved techniques for inferring both pose and shape from limited visual input.
- Introducing SAM 3D: a New Standard for 3D Object & Human Reconstruction from a Single ImageMeta · AI at Meta on YouTube ·
- Meta Meta AI
- Meta AI is developing technologies for 3D reconstruction, including human bodies, from single images, which directly involves pose and shape estimation.
- Human pose estimation
- Focuses solely on the location of body joints, not the ov
- Action recognition
- Aims to classify the activity a person is performing, rather
- Facial landmark detection
- Concentrates on points of the face, not the entire bo
The field emerged from computer vision research in the late 20th and early 21st centuries, with significant progress driven by academic benchmarks and competitions focused on human pose estimation. Source ↗
Use a mobile app that offers augmented reality try-on features for clothing. When you point your phone's camera at yourself, the app estimates your body shape and pose to show how the clothes would fit, demonstrating pose and shape estimation in action.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
- HeuristicDouble coding agent task completion with docs linksProviding coding agents with a documentation link significantly increases their task completion rate. · 2 min
Questions people ask
- What is the difference between pose and shape estimation?
- Pose estimation finds the location of body joints, like elbows and knees. Shape estimation determines the physical dimensions and form of the body, such as its width and depth.
- Can AI estimate pose and shape from a single photo?
- Yes, advanced AI models can now estimate human pose and shape from a single 2D image, though accuracy can vary.
First seen on Subvolts Nov 19, 2025. Updated Sep 15, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.