4D SignWordNet: Building a Comprehensive Video Dataset for Bengali Sign Word Recognition Model and Language Model

This dataset contains videos illustrating Bengali Sign Language. 6 Bangladeshi adults created this dataset with the specific goal of addressing the absence of a comprehensive video dataset for Bengali sign language. This dataset serves as a bridge to fill this gap by offering a rich resource to develop real-life expert systems. This dataset aims to include all aspect of practical sign language usage, addressing the critical need for variety. This dataset comprises two parts as word and sentences each of which includes two sections. The first section involves Raw Video Data and the second involves the Masking Data. Within the Masking Video Data, there are two types of videos: with background and without background. For the word part, each section and sub section contain 102 classes with approximately 300 videos per class, resulting in a total of approximately 30,600 videos in each section and approximately 91,800 videos overall. And for the sentence part, each section and sub section comprise 20 classes with approximately 100 videos per class, resulting in a total of approximately 2000 videos in each section and approximately 6000 videos overall. Notably, both sections of the dataset prioritize diversity by capturing videos in various backgrounds and lighting conditions, ensuring its suitability for real-world applications.