File size: 2,713 Bytes
4dd3d4b
 
 
 
 
 
 
 
 
fa0d0ba
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4dd3d4b
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
---
license: apache-2.0
task_categories:
- translation
language:
- th
pretty_name: ThaiSignVis
---
The dataset comprises over 1,000 hours of videos featuring live hand sign translations, accompanied by text transcripts from Thai parliamentary meetings. The data can be accessed at kaggle link [here][4]
# Dataset Generation
A total of 153 live videos were downloaded from the [TPchannel][1]. These videos were streamed between August 18, 2024, and June 4, 2023. The transcripts were automatically generated using YouTube's subtitle algorithm.
The videos were then post-processed by trimming and cropping, focusing only on segments where the sign language interpreter is detected.

# Data Structure
### no_labels_video/
This directory contains the full videos that could not be processed for transcription. These videos are unedited due to the inability to fetch accurate transcripts.  

### process_videos/
This directory contains the processed videos where transcripts are available.

### detection_results.csv
This file provides the detection results for the sign language interpreter:
- `second `:  The time index in seconds.
- `cls`:   The classification result, where -1 means the sign language interpreter is not detected, and 0 indicates that the interpreter is detected.
- `x`, `y`, `w`, and `h`: The coordinates (x, y) and dimensions (width, height) of the bounding box where the sign language interpreter is detected. A value of -1 indicates the interpreter is not found.

### transcript_window_{window_id}.csv  
This file contains the text transcript for each window segment:
- `text`: The text transcript of the video.
- `start`: The timestamp (in seconds) indicating when the text starts relative to the original video.
- `duration`: The duration (in seconds) of the transcript.
- `relative_strart`: The timestamp (in seconds) indicating when the text starts relative to process_video_{window_id}.mp4.

### transcript.json 
This file contains the text transcript for the entire video:
- `text`: The text transcript of the video.
- `start`: The timestamp (in seconds) indicating when the text starts.
- `duration`: The duration (in seconds) of the transcript.

### process_video_{window_id}.mp4
The video segments where the sign language interpreters are performing.

# Limitations and Considerations
Please note that this dataset is still in the early stages of development. If you encounter any issues, feel free to contact me.

# Contact
Github: [TOMTUNN][2]
Linkedin: [Thanawut-tim ][3]
Email: Earth34709@gmail.com

[1]:https://www.youtube.com/@TPchannel10
[2]:https://github.com/Tomtunn
[3]:https://www.linkedin.com/in/thanawut-tim
[4]:https://www.kaggle.com/datasets/thanawuttimpitak/thaisignvis