Update README.md

README.md

@@ -60,51 +60,104 @@ Key observations:
 
 ---
 
-## 📐 Data Format
+## 📁 Dataset Structure
+
+```
+dataset_root/
+├── PointCloud_Asset/
+├── Video_Asset/
+├── Label_Asset/
+├── PointCloud_Combine/
+│   ├── train/
+│   ├── val/
+│   └── test/
+├── Video_Combine/
+│   ├── train/
+│   ├── val/
+│   └── test/
+├── Label_Combine/
+│   ├── train/
+│   ├── val/
+│   └── test/
+└── meta/
+    ├── asset_meta.json
+    ├── combination_meta.json
+    ├── train_combination_meta.json
+    ├── val_combination_meta.json
+    ├── test_combination_meta.json
+    └── label_map.csv
+```
 
-Each sample in the dataset includes:
+---
 
-- `.ply` point clouds with per-vertex color
-- JSON-based metadata per asset with fields:
-  - `label_str`: class name  
-  - `gender`, `pose`, `type`  
-  - `wnlemmas`: fine-grained semantic tags
-- Segmentation masks (optional, task-dependent)
-- Textured mesh reconstructions (Blender + LCM pipeline)
+## 📦 Data Description
+
+### 🔹 Individual Asset Files
+
+- **PointCloud_Asset/**  
+  Contains raw point clouds of individual clothing or body parts in `.ply` format.
+
+- **Video_Asset/**  
+  Rendered 3D videos of individual assets showing different rotations or views.
+
+- **Label_Asset/**  
+  Label information (e.g., category, class ID) for each individual asset.
 
 ---
 
-## 🧪 Benchmarks
+### 🔹 Combined Assets (Mannequin Representations)
 
-Using **[SAMPart3D](https://yhyang-myron.github.io/SAMPart3D-website/)** as a baseline, our benchmark results show:
+- **PointCloud_Combine/**  
+  Combined point clouds representing mannequins wearing multiple assets. Split into `train`, `val`, and `test` sets.
 
-- **High segmentation accuracy** (mIoU: 0.9930)
-- **Class-dependent variation** in Average Precision (AP)
+- **Video_Combine/**  
+  Rendered 3D videos of mannequins with asset combinations. Also split into `train`, `val`, and `test`.
 
-![Result Visualization](figure/seg_tuning.png)
+- **Label_Combine/**  
+  Label files corresponding to the combined point clouds and videos.
 
 ---
 
-## 💡 Use Cases
+## 🗂️ Metadata Files (`meta/`)
+Each mata contains this detailed information:
+  - `label_str`: class name  
+  - `gender`, `pose`, `type`  
+  - `wnlemmas`: fine-grained semantic tags
+
+- **asset_meta.json**:  
+  Metadata for individual assets
+  
+- **combination_meta.json**:  
+  Metadata for all combinations
 
-The 3DFDReal dataset enables a wide range of applications, including:
+  
+- **train_combination_meta.json**, **val_combination_meta.json**, **test_combination_meta.json**:  
+  Define which combinations belong to each data split.
 
-- Virtual try-on and digital styling
-- Metaverse-ready asset generation
-- Pose-aware segmentation and label propagation
-- Avatar rigging and deformation modeling
+- **label_map.csv**:  
+  Maps label for the first data acquisition fullid and the second acquisition label fullid.
 
 ---
 
-## 📄 Citation
+## 📃 License
+
+CC-BY 4.0
+
+---
 
-If you use this dataset in your work, please cite:
+## 📚 Citation
 
 ```bibtex
-@inproceedings{lim2025volme,
+@misc{3DFDReal,
   title={3DFDReal: 3D Fashion Data from the Real World},
-  author={Jiyoun Lim, Jungwoo Son, Alex Lee, Sun-Joong Kim, Nam Kyung Lee, Won-Joo Park},
-  booktitle={TBD},
-  year={2025}
+  author={Jiyoun Lim, Jungwoo Son, Alex Lee, Sun-Joong Kim, Nam Kyung Lee, Won-Joo Park},,
+  year={2025},
+  howpublished={\url{https://huggingface.co/datasets/kusses/3DFDReal}},
 }
-```
+```
+
+---
+
+## 💬 Contact
+
+For questions, please reach out via [kusses@etri.re.kr] or use the Discussions tab on Hugging Face.