SSL Dataset

This dataset is the result of a collaborative effort between King Fahd University of Petroleum and Minerals (KFUPM) and the National Center for Artificial Intelligence (NCAI). It contains sign language data collected from a diverse group of signers, designed to support research and development in the field of sign language recognition and related AI applications.

Data Summary

The dataset includes sign language sentences from a variety of signers, capturing different signing styles and contexts. Below is a brief overview of the dataset:

Number of Signers: 18
Total Hours of Video: 35 hours
Data Splits:
- Training Split:
  - Includes three female signers and eleven male signers.
  - The total duration of the training split is 34 hours.
- Test Split:
  - The test split is 7 hours in total, consisting of:
    - Test 1: Unseen signers and unseen sentences.
    - Test 2: Seen signers and unseen sentences.
    - Test 3: Seen sentences and unseen signers.

The table below summarizes the data for each split based on the sentences, minutes, seen sentences, seen signers, and other parameters.

Styled Table

Split	Sentences	Minutes	Seen Sentences	Seen Signers	# Samples	# Signers	Gender
Train	24,111	2,017.82	Yes	Yes	1,900	16	4F, 12M
Test 1	200	16.65	No	No	100	2	1F, 1M
Test 2	1,297	107.95	No	Yes	100	11	3F, 10M
Test 3	3,783	337.33	Yes	No	1,900	2	1F, 1M

Data Access

Data Acess Requirement: This dataset can be accessed by clicking Kaggle dataset

Citation

When utilizing this dataset in your research, remember to cite the dataset paper once it is published.

license

Please follow the licensing terms of Creative Commons Attribution-NonCommercial (CC BY-NC).

Styled List