A Multimodal BISINDO Corpus: Annotated Images and Videos for Indonesian Sign Language Hand Gesture Recognition

The data were collected under two environmental conditions: indoor recordings (with both bright and low-light variations) and outdoor recordings. Each recording session produced video sequences that were subsequently extracted into image frames, followed by annotation and labeling according to the corresponding alphabet gesture.

In total, the extraction process yielded 19,760 image samples, capturing variations in lighting conditions, hand motion dynamics, and recording perspectives. However, due to ethical considerations and the need to protect participant identity, the publicly shared version of this dataset includes only the video data from a single actor, whose facial region has been anonymized through blurring to prevent personal identification while preserving the integrity of the gesture data.