Update README.md

README.md

@@ -14,24 +14,41 @@ size_categories:
 - 100B<n<1T
 ---
 
-Please visit our [Homepage](https://caizhongang.com/projects/GTA-Human/) and [Toolbox](https://github.com/caizhongang/gta-human_toolbox/tree/main) for more information.
+# Playing for 3D Human Recovery (TPAMI 2024)
+- [Homepage](https://caizhongang.com/projects/GTA-Human/)
+- [Toolbox](https://github.com/caizhongang/gta-human_toolbox/tree/main)
+- [Paper](https://arxiv.org/abs/2110.07588)
 
-Abstract: 
-Image- and video-based 3D human recovery ( i.e. , pose and shape estimation) have achieved substantial progress. 
-However, due to the prohibitive cost of motion capture, existing datasets are often limited in scale and diversity. 
-In this work, we obtain massive human sequences by playing the video game with automatically annotated 3D ground truths. 
-Specifically, we contribute GTA-Human, a large-scale 3D human dataset generated with the GTA-V game engine, featuring 
-a highly diverse set of subjects, actions, and scenarios. More importantly, we study the use of game-playing data 
-and obtain five major insights. First , game-playing data is surprisingly effective. A simple frame-based baseline 
-trained on GTA-Human outperforms more sophisticated methods by a large margin. For videobased methods, GTA-Human 
-is even on par with the in-domain training set. Second , we discover that synthetic data provides critical complements 
-to the real data that is typically collected indoor. We highlight that our investigation into domain gap provides 
-explanations for our data mixture strategies that are simple yet useful, which offers new insights to the research 
-community. Third , the scale of the dataset matters. The performance boost is closely related to the additional 
-data available. A systematic study on multiple key factors (such as camera angle and body pose) reveals that the 
-model performance is sensitive to data density. Fourth , the effectiveness of GTA-Human is also attributed to the 
-rich collection of strong supervision labels (SMPL parameters), which are otherwise expensive to acquire in real 
-datasets. Fifth , the benefits of synthetic data extend to larger models such as deeper convolutional neural networks 
-(CNNs) and Transformers, for which a significant impact is also observed. We hope our work could pave the way for 
-scaling up 3D human recovery to the real world.
+## Updates
 
+- [2024-10-02] GTA-Human datasets are now available on HuggingFace!
+- [2024-09-19] Release of GTA-Human II Dataset
+- [2022-07-08] Release of GTA-Human Dataset on MMHuman3D
+
+## Datasets
+
+Please click on the dataset name for download links and visualization instructions.
+
+| Features | [GTA-Human](https://caizhongang.com/projects/GTA-Human/gta-human.html) | [GTA-Human II](https://caizhongang.com/projects/GTA-Human/gta-human_v2.html) |
+| :------------------------ | :-------: | :-------: |
+| Num of Scenes             | 20,005    | 10,224    |
+| Num of Person Sequences   | 20,005    | 35,352    |
+| Color Images              | Yes       | Yes       |
+| 3D BBox & Point Cloud     | No        | Yes       |
+| Parametric Model          | SMPL      | SMPL-X    |
+| Num of Persons per Scene  | 1         | 1-6       |
+
+## Citation
+```text
+@ARTICLE{10652891,
+  author={Cai, Zhongang and Zhang, Mingyuan and Ren, Jiawei and Wei, Chen and Ren, Daxuan and Lin, Zhengyu and Zhao, Haiyu and Yang, Lei and Loy, Chen Change and Liu, Ziwei},
+  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
+  title={Playing for 3D Human Recovery}, 
+  year={2024},
+  volume={},
+  number={},
+  pages={1-12},
+  keywords={Three-dimensional displays;Annotations;Synthetic data;Shape;Training;Parametric statistics;Solid modeling;Human Pose and Shape Estimation;3D Human Recovery;Parametric Humans;Synthetic Data;Dataset},
+  doi={10.1109/TPAMI.2024.3450537}
+}
+```