"BdSL175: A Bangladeshi Sign Language Image Dataset"

"BdSL175 is a large-scale image dataset designed to support research and real-time applications in Bangladeshi Sign Language recognition. Despite more than 300 million Bangla speakers worldwide, the lack of a publicly available, large-scale BdSL image dataset has hindered progress in assistive communication technologies. BdSL175 addresses this gap by providing 35,000 high-quality RGB images across 175 commonly used Bangladeshi Sign Language word classes, with 200 samples per class to ensure class balance. The dataset was collected from five student volunteers aged 20-25 across multiple universities in Bangladesh using a diverse range of smartphones, from entry-level to flagship devices, to capture real-world variability in lighting, background and image quality. All images were manually inspected to ensure proper focus, framing and minimal occlusion, with ambiguous samples discarded or recollected. The dataset follows a Bangla-English naming convention for folders and metadata to support both local usability and international research interoperability. BdSL175 is organized in a ready-to-use format compatible with popular machine learning frameworks such as PyTorch or TensorFlow. This dataset is intended to serve as a foundational resource for sign language recognition, human-computer interaction research, mobile translation systems and accessibility-focused applications with scope for future expansion to include additional gestures, environments and video-based signing."