fhswf/wholebody-pose-estimation-fingerspelling

该数据集包含5,000张标注图像，用于微调全身姿势估计模型。图像内容为德国手语（Deutsche Gebärdensprache）手指字母的孤立手势。数据集的关键特征包括：5,000张标注图像、全身姿势估计任务（包括手、脸和身体关键点）、德国手语手指字母的孤立手势、MIT许可证、图像与对应的姿势标注文件（通常是JSON格式）。数据集旨在用于微调姿势估计模型、手语识别和翻译研究、开发辅助技术以及训练模型理解精细手势和全身动作。数据集的局限性包括仅包含孤立手势、光照和背景变化未系统控制，且特定于德国手指字母，可能无法直接推广到其他手语。

This dataset contains 5,000 annotated images for fine-tuning whole-body pose estimation models. The images depict individuals performing signs from the German Sign Language (Deutsche Gebärdensprache) finger alphabet. Key features include: 5,000 annotated images, whole-body pose estimation task (including hands, face, and body keypoints), isolated signs from the DGS finger alphabet, MIT license, and images with corresponding pose annotation files (typically JSON format). The dataset is designed for fine-tuning pose estimation models, research in sign language recognition and translation, developing assistive technologies, and training models to understand fine-grained hand shapes and full-body gestures. Limitations include the focus on isolated signs, variations in lighting and background, and specificity to the German finger alphabet, which may not generalize directly to other sign languages.