add compelling why chingmu section

README.md

@@ -29,6 +29,19 @@ dataset_info:
 # ChingMu 1000-Hour Embodied Motion Dataset
 ### 青瞳1000小时具身智能动作数据集
 
+## 🌟 Why ChingMu?
+
+**The largest and most precise optical motion capture dataset purpose-built for humanoid robots and dexterous manipulation.**
+
+Unlike monocular video-based datasets, ChingMu provides **sub-millimeter accuracy**, **120 fps temporal resolution**, and **co-registered multi-modal signals** — including full-body skeleton, finger articulation, object 6D pose, and synchronized multi-view video. Every take is manually quality-checked and robot-retargeted, eliminating the noise and ambiguity common in internet-sourced data.
+
+**Key differentiators:**
+- ✅ **Optical ground truth** – not estimated, not synthetic
+- ✅ **Dexterous hand data** – 20+ DoF per hand, synchronized with object tracking
+- ✅ **Robot-ready** – pre-retargeted to Unitree G1 and customizable to your platform
+- ✅ **Real-world scenarios** – 15+ environments, 500+ tasks, 1000+ objects
+- ✅ **Scalable** – from single-task samples to full 1000-hour corpus
+
 > High-precision **optical motion capture** data for humanoid robots, dexterous hands, embodied AI, and virtual production.
 
 | | |