Missing depth data
Viewpoint and occlusion prevent complete depth reconstruction.
Fine motor research for non-human primates
3D-AI
Fine Hand Tracking & Reconstruction
Markerless 3D reconstruction of 44 bilateral hand keypoints using six synchronized cameras and deep learning.
44KEYPOINTS
6CAMERAS
60 fpsMAX RATE
≤ 8 mmACCURACY
RESEARCH CHALLENGE
Fine movement should go beyond visual observation
Hand motion links behavior to neural mechanisms, yet conventional methods miss fast, subtle and occluded finger movements.
Manual scoring limits
Time-consuming, subjective and difficult to scale.
Markers interfere
Markers alter natural behavior and are unsuitable for long studies.
SOLUTION
From multi-view video to 3D motion data
An integrated pipeline spanning synchronized capture, AI detection, reconstruction, validation and export.
HAND PIPELINE · LIVE
›
44-POINT MODEL
A 44-point model covering the complete hand motion chain
Twenty-two standardized keypoints per hand cover the elbow, wrist and every finger segment for complete bilateral motion analysis.
3D RECONSTRUCTION / BI-MANUAL KEYPOINTS
Complete hand motion chain. 44-point fine analysis. Turn grasping and manipulation into measurable 3D data.
3D reconstruction demo
2Elbow & wrist
REFERENCE CHAIN / MOTION BASE
Establish the hand motion reference chain. Provide spatial baseline for fingers and wrist.
20Finger joints
FINE JOINTS / SEGMENT ANALYSIS
Four joint points on every finger. Cover flexion, aperture and manipulation from base to tip.
CORE CAPABILITIES
Built for real experimental settings
Every layer, from capture hardware to analysis, is designed for fine hand movement research.
Markerless 3D reconstruction
Preserve more natural behavior with no physical markers.
MARKERLESS
Six-camera synchronized capture
High-speed multi-angle coverage of the hand workspace.
6 × 60 FPS
Specialized keypoint model
Optimized for primate hand anatomy and fine movements.
44 POINTS
Automated pipeline
Automatic processing from video to coordinates and visualization.
Custom model training
Adapt to species, subjects, lighting, cages and tasks.
Validation and filtering
Sub-centimeter accuracy validation plus optional Kalman filtering.
SYSTEM PERFORMANCE
Key specifications at a glance
SUBJECTPrimate handsbilateral
KEYPOINTS4422 per hand
CAMERAS6synchronized
FRAME RATE60 fpsmax
RESOLUTION1280 × 1024pixels
ACCURACY≤ 8 mmmedian error
Flexible deployment
SERVERLOCALSupports server deployment and local deployment based on data security, compute and collaboration needs.
DATA OUTPUT
Turn every hand movement into computable data
Output files preserve the complete chain from 2D detection to 3D reconstruction for review, visualization and analysis.
2D keypoint sequencesPer-camera detection results
2D CSV2D tracking videosQuick quality review
VIDEO3D coordinate sequences44-point standardized motion data
3D CSV3D reconstruction videoVisualize complete movement
VIDEOAccuracy reportQuantified sub-cm reconstruction accuracy
REPORTKINEMATIC ANALYSIS
TrajectoryJoint angleSpeed / AccelerationHand apertureBilateral coordinationMovement stability
Metrics can be derived from 3D coordinate sequencesRESEARCH APPLICATIONS
Connect fine behavior, neural mechanisms and intervention assessment
Fine motor control
Study regulation across fingers, wrist and forearm.
Grasp & manipulation
Quantify strategies across objects and task difficulty.
Neural circuits
Align with electrophysiology, imaging or neuromodulation.
Disease model assessment
Analyze motor deficits in Parkinson’s and spinal injury models.
Intervention evaluation
Compare fine movement before and after interventions.
Tool use & cognition
Study bimanual coordination and goal-directed behavior.
CUSTOM SOLUTION
A dedicated solution built around your paradigm
Configure the full solution around your site, species, model, interfaces and deployment needs.
01
Site survey and six-camera layout
02
Species, subject and environment adaptation
03
Customer-data annotation and model training
04
Custom keypoints and skeleton topology
05
Device synchronization and data interfaces
06
Server and local deployment configuration
DiscoveryAssessmentDesignPilotTrainingDelivery
Build a shareable, comparable and reproducible hand-motion data foundation
Unified keypoint definitions and 3D coordinates establish a standardized base for cross-study comparison, review and future databases.