---
title: Pose Estimation
description: >-
  What pose estimation is, how computer vision detects body joints and landmarks
  in images, and why it underpins virtual try-on and AR experiences.
shortDefinition: >-
  Computer vision technology that detects the position of human body joints and
  landmarks in images or video.
relatedTerms:
  - image-segmentation
  - garment-transfer
  - ar-try-on
  - 3d-body-model
heroImage: /glossary-images/pose-estimation-skeleton.webp
keywords:
  - pose estimation
  - body tracking
  - keypoint detection
publishedAt: '2026-06-10'
updatedAt: '2026-06-10'
noindex: false
canonicalPath: /glossary/pose-estimation
---

## What is pose estimation?

**Pose estimation** is a computer vision technique that identifies the position of human body joints (shoulders, elbows, wrists, hips, knees, ankles) in an image or video stream. The output is a set of keypoint coordinates, usually connected into a skeleton that describes the person's posture.

In try-on systems, pose estimation answers the question *how is this body positioned?* That answer determines where a garment lands, at what angle, and at what scale.

![A detected body skeleton overlaid on a photo](/glossary-images/pose-estimation-skeleton.webp)

## How it works

1. **Person detection**: locate the human subject in the frame.
2. **Keypoint prediction**: a neural network predicts the coordinates of 17–33 body joints.
3. **Skeleton assembly**: keypoints are connected into a structure tracking posture and limb angles.
4. **Temporal smoothing** (video only): predictions are smoothed across frames to reduce jitter.

## Role in virtual try-on

- **Image-based try-on**: a single-photo pose estimate guides [garment transfer](/glossary/garment-transfer.md), so sleeves follow the arms and the hem falls at the right height.
- **Live [AR try-on](/glossary/ar-try-on.md)**: per-frame pose estimation anchors 3D products (shoes, watches) to the moving body.
- **Body measurement**: joint positions help estimate proportions for [3D body models](/glossary/3d-body-model.md) and size recommendation.

Inaccurate pose estimation produces misaligned results: a dress draped at the wrong angle or shoes offset from the feet.

<Faq>
  <FaqItem question="Is pose estimation the same as body scanning?">
    No. Pose estimation detects joint positions in 2D (or 2.5D) from ordinary images. Body scanning builds a full 3D mesh of the body surface, which requires specialized hardware or multi-angle capture.
  </FaqItem>
</Faq>