caizhongang
commited on
Commit
•
69b0b50
1
Parent(s):
0e40f5a
Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,30 @@
|
|
1 |
-
---
|
2 |
-
license: other
|
3 |
-
license_name: s-lab-license
|
4 |
-
license_link: LICENSE
|
5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
license_name: s-lab-license
|
4 |
+
license_link: LICENSE
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
tags:
|
8 |
+
- Synthetic Data
|
9 |
+
- 3D Humans
|
10 |
+
- Human Pose and Shape Estimation
|
11 |
+
- Human NeRF
|
12 |
+
pretty_name: SynBody
|
13 |
+
size_categories:
|
14 |
+
- 100B<n<1T
|
15 |
+
---
|
16 |
+
|
17 |
+
# SynBody: Synthetic Dataset with Layered Human Models for 3D Human Perception and Modeling
|
18 |
+
|
19 |
+
[Homepage](https://synbody.github.io/)
|
20 |
+
|
21 |
+
Abstract: Synthetic data has emerged as a promising source for 3D human research as it offers low-cost access to large-scale human datasets.
|
22 |
+
To advance the diversity and annotation quality of human models, we introduce a new synthetic dataset, SynBody, with three appealing features:
|
23 |
+
1) a clothed parametric human model that can generate a diverse range of subjects;
|
24 |
+
2) the layered human representation that naturally offers high-quality 3D annotations to support multiple tasks;
|
25 |
+
3) a scalable system for producing realistic data to facilitate real-world tasks.
|
26 |
+
|
27 |
+
The dataset comprises 1.2M images with corresponding accurate 3D annotations, covering 10,000 human body models, 1,187 actions, and various viewpoints.
|
28 |
+
The dataset includes two subsets for human pose and shape estimation as well as human neural rendering. Extensive experiments on SynBody indicate that
|
29 |
+
it substantially enhances both SMPL and SMPL-X estimation. Furthermore, the incorporation of layered annotations offers a valuable training resource
|
30 |
+
for investigating the Human Neural Radiance Fields (NeRF).
|