Nigerian Sign Language (NSL) Alphanumeric Dataset

The Nigerian Sign Language (NSL) Alphanumeric Dataset is an annotated computer vision dataset created to benchmark and deploy lightweight deep learning models for real-time sign recognition. The dataset covers 33 static alphanumeric sign gestures, including numbers (0–9) and alphabet letters (A–Y, excluding motion-based signs J and Z, and composite gesture 10). Data was collected under varied indoor/outdoor lighting conditions, camera angles, and background environments. Scope & Exclusions- Included Classes (33 Total): Numbers 0–9; Letters A–Y (excluding J and Z).- Excluded Classes: J and Z (motion-dependent) and 10. Dataset StructurePre-split into train, val, and test directories following standard YOLO formatting:- images/: .jpg gesture frames across multiple subjects.- labels/: Matching normalized YOLO .txt files.- dataset.yaml: Class mappings defining indices 0 through 32. Intended Use Cases- Benchmarking lightweight real-time object detectors (e.g., YOLOv8, YOLO11).- Assistive technology research for deaf and hard-of-hearing communities.- Edge device deployment (Android/iOS/ONNX/TFLite). Comprehensive alphanumeric sign gesture dataset split into train, validation, and test sets for object detection models.