Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,68 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: odbl
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: odbl
|
| 3 |
+
Author: Dexrobot Team
|
| 4 |
+
---
|
| 5 |
+
# MOCAP
|
| 6 |
+
## - Human Dexterous Manipulation Dataset & Simulation Platform
|
| 7 |
+
|
| 8 |
+
## 🔬 Coming Soon - Expected Release: Mid-October 2025
|
| 9 |
+
|
| 10 |
+
We are excited to announce the upcoming release of MOCAP, a comprehensive dataset and simulation platform for human dexterous manipulation research.
|
| 11 |
+
|
| 12 |
+
📊 Dataset Overview
|
| 13 |
+
|
| 14 |
+
Our team has collected and processed approximately 10,000 recordings / 200 hours of human manipulation data, categorized by:
|
| 15 |
+
|
| 16 |
+
- 33 Common Grasping Gestures: Comprehensive coverage of human grasping patterns
|
| 17 |
+
- ~10 In-Hand Manipulation Actions: Essential dexterous manipulation skills
|
| 18 |
+
|
| 19 |
+
🛠️ Key Features
|
| 20 |
+
|
| 21 |
+
Data Pipeline
|
| 22 |
+
|
| 23 |
+
- ✅ Complete data cleaning and preprocessing pipeline
|
| 24 |
+
- ✅ Quality-assured human operation recordings
|
| 25 |
+
- ✅ Standardized annotation framework
|
| 26 |
+
|
| 27 |
+
Real2Sim Achievement
|
| 28 |
+
|
| 29 |
+
- 🎯 Precise replication of human hand operations in simulation environment
|
| 30 |
+
- 🤖 Reinforcement Learning-based real2sim conversion methodology
|
| 31 |
+
- 📈 100x data scaling through simulation synthesis
|
| 32 |
+
|
| 33 |
+
🎯 Technical Significance
|
| 34 |
+
|
| 35 |
+
1. End-to-End Validation
|
| 36 |
+
|
| 37 |
+
Complete workflow from "how to collect" to "how to use", ensuring data collection and preprocessing accuracy.
|
| 38 |
+
|
| 39 |
+
2. Physics-Informed Insights
|
| 40 |
+
|
| 41 |
+
Obtain contact points and force information through physical simulation - data difficult to observe directly in real-world scenarios.
|
| 42 |
+
|
| 43 |
+
3. Integrated Infrastructure
|
| 44 |
+
|
| 45 |
+
Build a unified collection-simulation-training infrastructure, preparing for scaling law applications through synthetic data generation.
|
| 46 |
+
|
| 47 |
+
🚀 What's Coming
|
| 48 |
+
|
| 49 |
+
- 📦 Complete Dataset Release: 10K+ human manipulation recordings
|
| 50 |
+
- 🔧 Preprocessing Tools: Ready-to-use data pipeline
|
| 51 |
+
- 🎮 Simulation Environment: Physics-based replication platform
|
| 52 |
+
- 📚 Documentation & Tutorials: Comprehensive usage guides
|
| 53 |
+
|
| 54 |
+
📅 Release Timeline
|
| 55 |
+
|
| 56 |
+
Expected Release Date: Mid-October 2025
|
| 57 |
+
|
| 58 |
+
Stay tuned for updates!
|
| 59 |
+
|
| 60 |
+
---
|
| 61 |
+
Developed by DexGym TeamLast Updated: September 11, 2025
|
| 62 |
+
|
| 63 |
+
📬 Contact
|
| 64 |
+
|
| 65 |
+
For early access requests or collaboration inquiries, please reach out to our team.
|
| 66 |
+
|
| 67 |
+
---
|
| 68 |
+
⭐ Star this space to get notified when the dataset becomes available!
|