Scene-PHOENIX
收藏arXiv2022-11-01 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2211.00448v1
下载链接
链接失效反馈官方服务：
资源简介：
Scene-PHOENIX是一个专为连续手语识别（CSLR）模型设计的背景鲁棒性基准数据集，由韩国科学技术院（KAIST）创建。该数据集通过利用现有的CSLR基准数据集，自动合成包含多样背景的手语视频，模拟真实世界环境，以评估模型在背景变化下的鲁棒性。数据集包含629个样本，背景图像来自LSUN和SUN397数据集，每个场景类在数据集中均匀分布。Scene-PHOENIX旨在解决CSLR模型在非工作室背景下的识别问题，提高模型的实际应用能力。
Scene-PHOENIX is a background-robust benchmark dataset specifically designed for continuous sign language recognition (CSLR) models, created by the Korea Advanced Institute of Science and Technology (KAIST). This dataset automatically synthesizes sign language videos with diverse backgrounds by leveraging existing CSLR benchmark datasets, simulating real-world scenarios to evaluate model robustness against background variations. It contains 629 samples, with background images sourced from the LSUN and SUN397 datasets, and each scene category is uniformly distributed across the dataset. Scene-PHOENIX aims to address the recognition challenges of CSLR models in non-studio environments and enhance their real-world application capabilities.
提供机构：
韩国科学技术院 (KAIST)
创建时间：
2022-11-01
