21-Keypoint Skeletal Hand Tracking
Precise 21-keypoint 3D finger joint coordinates for dexterous hand manipulation, pinch force estimation, and imitation learning robotics.
Quality & Precision Benchmarks
3D JOINT ERROR
< 2.4 mm
SKELETAL NODES
21 Joints per Hand
TRACKING RATE
60 Hz Continuous
FORMAT STANDARD
LeRobot / MANO Compatible
Dataset Taxonomy & Output Structure
wrist_joint_3d (x, y, z in mm)
thumb_mcp_pip_dip_tip_3d (4 Nodes)
index_mcp_pip_dip_tip_3d (4 Nodes)
middle_ring_pinky_joints_3d (12 Nodes)
Specific Type Tasks & Applications
- • Dexterous Humanoid Gripper Imitation
- • Fine Motor Skill Assembly Automation
- • Hand-Object Interaction Synthesis
What Is Right vs What Is Wrong
| COMMON COMPETITOR ERRORS (WRONG) | BLUE PROJECTS GROUND TRUTH (RIGHT) |
|---|---|
| FAIL: 2D bounding boxes on hands lacking 3D finger joint depth | PASS: 21 sub-millimeter 3D skeletal finger joint coordinates |
| FAIL: Occluded finger joint dropout causing robotic execution jitter | PASS: Multi-camera camera-glove fusion maintaining continuous joint tracking |
Files & Telemetry Data Example (Python)
import json
import numpy as np
# Load Blue Projects Egocentric Data Type: 21-Keypoint Skeletal Hand Tracking
with open("21-keypoint-skeletal-hand-tracking_sample.json", "r") as f:
data = json.load(f)
print("Loaded Frame Keys:", list(data.keys()))
Why Blue Projects for 21-Keypoint Skeletal Hand Tracking?
Request a free matched 10-sequence sample batch formatted to your exact VLA foundation model requirements.
Request Free Sample Batch →