ASLAD-190K

The ASLAD-190K dataset is a comprehensive and carefully designed collection featuring 190,000 labeled RGB images of 32 Arabic Sign Language (ArSL) alphabets. It has been specifically developed to support machine learning and computer vision research, making it an indispensable resource for training models in sign language recognition and alphabet classification.

Key Features of the ASLAD-190K Dataset

1. Extensive Image Collection:
   The dataset provides 190,000 high-quality RGB images, each representing an alphabet of the Arabic Sign Language. These images are clearly labeled, which simplifies training and testing for various AI applications.

2. Diverse Image Captures:
   Images were captured using two webcams, the HP HD Camera and HP Truevision HD, with the MediaPipe library ensuring precise and consistent results. Moreover, the data collection process included variations to make the dataset highly adaptable:

   • Lighting Conditions: Photos were taken in both bright and dim lighting.
   • Distances: Images range from close-up to far zoom levels.
   • Timing: Both daytime and nighttime captures were included.
   • Rotations: 2D and 3D rotations simulate various hand orientations.
   • Backgrounds: A variety of settings were used for greater contextual richness.

3. Real-World Simulation:
   To maximize usability in real-world applications, the dataset emphasizes diversity. For example, changes in lighting, distance, and orientation were intentionally introduced, which prepares AI models to handle complex, real-life scenarios effectively.

4. Augmentation for Enhanced Performance:
   Extensive augmentation techniques were applied to the dataset, which significantly improve its robustness. Consequently, this ensures better performance of machine learning models during training and testing, even under intricate conditions.