Create README.md

README.md

@@ -0,0 +1,23 @@
+# EgoDynamic4D Dataset from AAAI 2026 paper: [Understanding Dynamic Scenes in Egocentric 4D Point Clouds](https://arxiv.org/abs/2508.07251)
+
+🚀 **EgoDynamic4D dataset is coming soon.**
+
+This repository hosts the official release of **EgoDynamic4D**, a large-scale egocentric **4D dynamic scene understanding** benchmark introduced in our AAAI 2026 paper:
+
+> **Understanding Dynamic Scenes in Egocentric 4D Point Clouds**
+
+---
+
+## About the Dataset
+
+**EgoDynamic4D** is a question answering (QA) benchmark designed for fine-grained **spatio-temporal reasoning** in egocentric dynamic scenes.
+
+The dataset includes:
+
+* Egocentric RGB-D videos
+* Camera poses
+* Globally unique instance masks
+* 4D bounding boxes over time
+* Large-scale QA annotations for dynamic reasoning tasks
+
+The dataset is constructed based on existing egocentric 4D resources, building upon **ADT and THUD++**. Our main contribution lies in the **large-scale, task-driven QA annotations**, which enable fine-grained spatio-temporal reasoning in dynamic egocentric scenes.