Search papers, labs, and topics across Lattice.
This paper introduces Bengal-HP_RU, the first head pose dataset focused on Bengali individuals, addressing the significant underrepresentation of South Asian populations in existing datasets. Comprising 12,894 labeled head images with continuous yaw, pitch, and roll annotations, the dataset was meticulously curated from Wikimedia Commons and includes diverse subjects across various demographics and conditions. The careful partitioning by uploader identity ensures data integrity, providing a robust resource for advancing head pose estimation research in underrepresented populations.
The Bengal-HP_RU dataset reveals the critical gap in head pose research, showcasing the rich diversity of Bengali individuals that has been largely overlooked.
Existing head pose datasets predominantly feature subjects of Western or East Asian origin, leaving South Asian populations, particularly Bengali individuals, largely underrepresented. We introduce Bengal-HP_RU, the first publicly available head pose dataset centred on Bengali subjects, comprising 12,894 labelled head images annotated with continuous yaw, pitch, and roll values. Images were collected from Wikimedia Commons under free licences and processed through an automated pipeline followed by manual label correction. The dataset is partitioned by Wikimedia uploader identity to prevent data contamination, yielding 10,494 training and 2,400 test images across 296 unique uploaders. Bengal-HP_RU exhibits substantial diversity in subject age, gender, occlusion, illumination, and background, reflecting realistic in-the-wild conditions. The dataset is publicly available at https://doi.org/10.17632/xbw9kr37jb.2.
