Sign-Language Datasets at Scale: A Comprehensive Survey on Resources, Benchmarks, and Annotation Standards

The article presents a comprehensive survey of sign-language datasets, indexing 120 resources across 35 sign languages and addressing challenges like modality imbalance and annotation granularity. It introduces a 24-field Sign-Language Datasheet and a public GitHub repository for standardized documentation, aiming to enhance the development of robust sign-language technologies. This work is significant for practitioners as it provides a unified framework for dataset design and evaluation, facilitating better training and evaluation of sign-language recognition and translation systems.