Kateht commited on
Commit
9686a7c
·
verified ·
1 Parent(s): dc0f09e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +44 -36
README.md CHANGED
@@ -1,59 +1,62 @@
1
  ---
2
- language: vi
3
  license: mit
4
  tags:
5
- - sign-language
6
- - mediapipe
7
- - vietnamese
8
- - keypoints
9
- - npz
10
  size_categories:
11
- - 50GB+
12
  task_categories:
13
- - sign-language-recognition
 
14
  pretty_name: VOYA Vietnamese Sign Language (VOYA_VSL)
15
  ---
16
 
17
  # VOYA Vietnamese Sign Language (VOYA_VSL)
18
 
19
- VOYA_VSL bộ dữ liệu ngôn ngữ hiệu tiếng Việt, được thu thập và tiền xử lý để phục vụ nghiên cứu về **nhận dạng ngôn ngữ ký hiệu**.
20
- Bộ dữ liệu bao gồm các chuỗi keypoints được trích xuất từ MediaPipe, mỗi chuỗi biểu diễn một video hiệu.
21
 
22
- ## 📦 Thông tin dữ liệu
23
 
24
- - **Định dạng**: `.npz` (NumPy compressed)
25
- - **Số mẫu**: ~161
26
- - **Cấu trúc một sample**:
27
- - `sequences`: ma trận shape `(60, 1605)` (60 khung hình, 1605 keypoints/feature mỗi khung)
28
- - `labels`: chỉ số lớp (int32)
29
 
30
- ## 📂 Cấu trúc dataset
31
 
 
32
  VOYA_VSL/
33
  │── Merged/
34
- ├── class_0001.npz
35
- ├── class_0002.npz
36
- ├── ...
37
- └── class_0161.npz
38
  │── labels.json
39
  │── dataset.py
40
  └── README.md
 
41
 
42
- - `Merged/` : chứa dữ liệu từng lớp đã gom thành `.npz`
43
- - `labels.json` : ánh xạ `class_xxxx` → tên ký hiệu
44
- - `dataset.py` : script để load dataset qua Hugging Face Datasets
45
- - `README.md` : tả dataset
46
-
47
- ## 🛠️ Cách sử dụng
48
 
 
49
  ```python
50
  from datasets import load_dataset
51
 
52
- # Load dataset VOYA_VSL
53
  dataset = load_dataset("Kateht/VOYA_VSL")
54
 
55
  print(dataset)
56
 
 
57
  DatasetDict({
58
  train: Dataset({
59
  features: ['sequences', 'labels'],
@@ -61,31 +64,36 @@ DatasetDict({
61
  })
62
  })
63
 
64
- # Truy cập 1 sample
65
  sample = dataset["train"][0]
66
  print(sample["sequences"].shape) # (60, 1605)
67
  print(sample["labels"]) # int label
68
 
69
- 📌 Bạn có thể tự chia train/val/test sau khi load:
70
-
71
  dataset = dataset["train"].train_test_split(test_size=0.2, seed=42)
 
72
 
73
- 🔖 Labels
74
- File labels.json chứa mapping từ ID lớp → nhãn tiếng Việt. Ví dụ:
75
 
 
 
76
  {
77
  "class_0001": "địa chỉ (bắc)",
78
  "class_0002": "địa chỉ (nam)",
79
  "class_0003": "địa chỉ (trung)"
80
  }
 
81
 
82
- 📜 Trích dẫn
83
- Nếu bạn sử dụng dataset này trong nghiên cứu, vui lòng trích dẫn:
84
 
 
 
 
85
  @misc{voya_vsl_2025,
86
  author = {Kateht et al.},
87
  title = {VOYA Vietnamese Sign Language Dataset},
88
  year = {2025},
89
  publisher = {Hugging Face},
90
  howpublished = {\url{https://huggingface.co/datasets/Kateht/VOYA_VSL}}
91
- }
 
 
1
  ---
2
+ language: en
3
  license: mit
4
  tags:
5
+ - sign-language
6
+ - mediapipe
7
+ - vietnamese
8
+ - keypoints
9
+ - npz
10
  size_categories:
11
+ - 50GB+
12
  task_categories:
13
+ - other
14
+ custom_task: sign-language-recognition
15
  pretty_name: VOYA Vietnamese Sign Language (VOYA_VSL)
16
  ---
17
 
18
  # VOYA Vietnamese Sign Language (VOYA_VSL)
19
 
20
+ VOYA_VSL is a Vietnamese Sign Language dataset collected and preprocessed for research in **sign language recognition**.
21
+ The dataset consists of sequences of keypoints extracted using MediaPipe, where each sequence represents a sign language video.
22
 
23
+ ## 📦 Dataset Information
24
 
25
+ - **Format**: `.npz` (NumPy compressed)
26
+ - **Number of samples**: ~161
27
+ - **Sample structure**:
28
+ - `sequences`: a matrix with shape `(60, 1605)` (60 frames, 1605 keypoints/features per frame)
29
+ - `labels`: class index (int32)
30
 
31
+ ## 📂 Dataset Structure
32
 
33
+ ```
34
  VOYA_VSL/
35
  │── Merged/
36
+ ├── class_0001.npz
37
+ ├── class_0002.npz
38
+ ├── ...
39
+ └── class_0161.npz
40
  │── labels.json
41
  │── dataset.py
42
  └── README.md
43
+ ```
44
 
45
+ - `Merged/` : contains per-class data stored as `.npz` files
46
+ - `labels.json` : maps `class_xxxx` → sign name
47
+ - `dataset.py` : script to load dataset with Hugging Face Datasets
48
+ - `README.md` : dataset description
 
 
49
 
50
+ ## 🛠️ Usage
51
  ```python
52
  from datasets import load_dataset
53
 
54
+ # Load the VOYA_VSL dataset
55
  dataset = load_dataset("Kateht/VOYA_VSL")
56
 
57
  print(dataset)
58
 
59
+ # Example output
60
  DatasetDict({
61
  train: Dataset({
62
  features: ['sequences', 'labels'],
 
64
  })
65
  })
66
 
67
+ # Access a sample
68
  sample = dataset["train"][0]
69
  print(sample["sequences"].shape) # (60, 1605)
70
  print(sample["labels"]) # int label
71
 
72
+ # Split train/validation/test after loading
 
73
  dataset = dataset["train"].train_test_split(test_size=0.2, seed=42)
74
+ ```
75
 
76
+ ## 🔖 Labels
 
77
 
78
+ The labels.json file maps class IDs to Vietnamese labels. Example:
79
+ ```
80
  {
81
  "class_0001": "địa chỉ (bắc)",
82
  "class_0002": "địa chỉ (nam)",
83
  "class_0003": "địa chỉ (trung)"
84
  }
85
+ ```
86
 
87
+ ## 📜 Citation
 
88
 
89
+ If you use this dataset in your research, please cite:
90
+
91
+ ```
92
  @misc{voya_vsl_2025,
93
  author = {Kateht et al.},
94
  title = {VOYA Vietnamese Sign Language Dataset},
95
  year = {2025},
96
  publisher = {Hugging Face},
97
  howpublished = {\url{https://huggingface.co/datasets/Kateht/VOYA_VSL}}
98
+ }
99
+ ```