ASLAD-190K: Arabic Sign Language Alphabet Dataset consisting of 190,000 Images

The ASLAD-190K dataset is an extensive collection containing 190,000 meticulously labeled RGB images representing 32 alphabets of the ArSL. To capture these images, we enlisted the help of two signers and utilized two different computer webcams, namely the HP HD camera and HP Truevision HD. The MediaPipe library was crucial in capturing RGB photos of various sizes.

During the data collection process, we took great care to introduce diversity into the images. This involved varying lighting conditions, distances (zoom in/out), timings (day and night), 2D and 3D image rotations, and backgrounds. As a result, the dataset now offers a significantly augmented image corpus that closely resembles real-world scenarios, encompassing increasingly intricate situations. This augmentation enhances the performance of classification models during training.