yonful commited on
Commit
1ef42da
·
verified ·
1 Parent(s): 9c0956c

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +70 -14
README.md CHANGED
@@ -5,41 +5,97 @@ task_categories:
5
  tags:
6
  - imu
7
  - activity-recognition
8
- - fitness
9
- - deep-learning
10
  size_categories:
11
- - 1M<n<10M
12
  ---
13
 
14
  # gem-analysis-humanml3d
15
 
16
- HumanML3D dataset with UniMTS augmented text descriptions for IMU pretraining
17
 
18
- ## 数据集信息
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19
 
20
- - **来源路径**: `data/HumanML3D`
21
- - **数据大小**: 1.0 GB
22
- - **用途**: 健身动作识别模型训练
23
 
24
- ## 使用方法
25
 
26
  ```python
27
  from huggingface_hub import snapshot_download
28
 
29
- # 下载数据集
30
  snapshot_download(
31
  repo_id="yonful/gem-analysis-humanml3d",
32
  repo_type="dataset",
33
- local_dir="./data/HumanML3D"
34
  )
35
  ```
36
 
37
- 或使用项目中的下载脚本:
38
 
39
- ```bash
40
- python scripts/prepare_data.py --dataset humanml3d
 
 
 
 
 
 
 
 
 
 
41
  ```
42
 
 
 
 
 
 
 
43
  ## 许可证
44
 
45
  请参考原始数据源的许可证要求。
 
 
 
 
 
 
 
 
 
 
 
 
 
5
  tags:
6
  - imu
7
  - activity-recognition
8
+ - motion-capture
9
+ - human-motion
10
  size_categories:
11
+ - 100K<n<1M
12
  ---
13
 
14
  # gem-analysis-humanml3d
15
 
16
+ HumanML3D 数据集,包含 3D 人体运动数据和文本描述。
17
 
18
+ ## 下载方式
19
+
20
+ ### 方式 1: 下载压缩版 (推荐,避免 rate limit)
21
+
22
+ 数据集包含 16 万+ 小文件,直接下载可能触发 rate limit。推荐下载压缩版:
23
+
24
+ ```bash
25
+ # 使用 huggingface-cli
26
+ huggingface-cli download --repo-type dataset yonful/gem-analysis-humanml3d humanml3d.tar.gz --local-dir .
27
+
28
+ # 解压
29
+ tar -xzf humanml3d.tar.gz
30
+ ```
31
+
32
+ 或使用 Python:
33
+
34
+ ```python
35
+ from huggingface_hub import hf_hub_download
36
+
37
+ # 下载压缩文件
38
+ hf_hub_download(
39
+ repo_id="yonful/gem-analysis-humanml3d",
40
+ filename="humanml3d.tar.gz",
41
+ repo_type="dataset",
42
+ local_dir="."
43
+ )
44
+
45
+ # 解压
46
+ import tarfile
47
+ with tarfile.open("humanml3d.tar.gz", "r:gz") as tar:
48
+ tar.extractall()
49
+ ```
50
 
51
+ ### 方式 2: 直接下载所有文件
 
 
52
 
53
+ 如果网络稳定,可以直接下载:
54
 
55
  ```python
56
  from huggingface_hub import snapshot_download
57
 
 
58
  snapshot_download(
59
  repo_id="yonful/gem-analysis-humanml3d",
60
  repo_type="dataset",
61
+ local_dir="./HumanML3D"
62
  )
63
  ```
64
 
65
+ ## 数据集结构
66
 
67
+ ```
68
+ HumanML3D/
69
+ ├── new_joints/ # 3D 关节位置数据
70
+ ├── new_joint_vecs/ # 旋转不变特征
71
+ ├── texts/ # 文本描述
72
+ ├── Mean.npy # 均值
73
+ ├── Std.npy # 标准差
74
+ ├── all.txt # 所有样本列表
75
+ ├── train.txt # 训练集
76
+ ├── val.txt # 验证集
77
+ ├── test.txt # 测试集
78
+ └── train_val.txt # 训练+验证集
79
  ```
80
 
81
+ ## 数据集信息
82
+
83
+ - 文件数量: ~160,000
84
+ - 原始大小: ~1.6 GB
85
+ - 压缩后大小: ~1.3 GB
86
+
87
  ## 许可证
88
 
89
  请参考原始数据源的许可证要求。
90
+
91
+ ## 引用
92
+
93
+ ```bibtex
94
+ @inproceedings{guo2022generating,
95
+ title={Generating diverse and natural 3d human motions from text},
96
+ author={Guo, Chuan and Zou, Shihao and Zuo, Xinxin and Wang, Sen and Ji, Wei and Li, Xingyu and Cheng, Li},
97
+ booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
98
+ pages={5152--5161},
99
+ year={2022}
100
+ }
101
+ ```