Real-Time 3D Skeleton Detection

Real-Time Measurement, Inference, and Analysis of Human Skeletons with AI Pose Estimation

AI Pose Estimation is an AI technology that estimates the posture of humans or animals from videos or images without the need of markers. VisionPose enables real-time skeleton detection with depth information using only a web camera, eliminating the need for depth sensors equipped cameras such as infrared. Additionally, it supports not only live camera footage but also still images and videos.

AI Pose Estimation ‘VisionPose’ Promotional Video

VisionPose is an AI Pose Estimation Engine that enables 2D/3D analysis of human skeletal and posture information without the need of marker. It supports not only live camera footage but also still images and videos.

VisionPose Single3D: Detects Human Skeletal Information with a Single Camera

Compatible with Unity and iOS/iPadOS! Markerless Easy Motion Capture. An AI Pose Estimation Engine Capable of Detecting Skeletons with Just One Camera.

Displays Only Skeletal Information Without Showing Human Body

VisionPose has a feature that analyzes and displays only the skeletal information of individuals captured by the camera. This feature allows for the analysis of people and their behavior without identifying individuals, making it suitable for situations where privacy is important.

Includes Convenient Skeleton Detection Tools for Analysis

VisionPose includes the following applications, enabling immediate use right after purchase.

  1. BodyAndColor: A sample app that enables real-time human skeletal information from webcam footage
  2. VP Analyzer: An app that enables skeleton detection from videos and still images

Common Drawbacks of Cameras with Infrared Sensors

While infrared depth sensors have several drawbacks (as of 2018 research), VisionPose has addressed these challenges, enabling support for various use cases that were previously difficult to achieve.

  1. Limited to front view, causing misidentification of back as front.
  2. Left-right reversal when switching from front to back, causing skeletal distortion when viewed from the side.
  3. Unable to capture skeletal data from above.
  4. Recognition based solely on infrared depth maps (without color) makes it difficult to distinguish between obstacles from people, resulting in weak performance for actions involving obscured body parts or holding objects.
  5. Struggles with outdoor recognition due to infrared limitations.
  6. Sensitivity to clothing or materials that absorb infrared light.

Back to Feature

Download PDF