GazeSeg
收藏资源简介:
GazeSeg数据集是由合肥工业大学和浙江大学联合创建的,旨在解决自然场景中视线跟随和目标识别的问题。该数据集包含77.5万张图像,每张图像都有像素级的注释,涵盖270个不同的目标类别。数据集的创建过程结合了现有的GazeFollow数据集,并通过引入新的注释协议来提高数据质量。GazeSeg数据集的应用领域广泛,包括社交互动、自闭症诊断和人与计算机交互等,旨在通过精确的像素级预测来提升视线跟随任务的性能。
GazeSeg is a dataset jointly created by Hefei University of Technology and Zhejiang University, which aims to address the challenges of gaze following and target recognition in natural scenes. This dataset consists of 775,000 images, each paired with pixel-level annotations covering 270 distinct target categories. During its development, GazeSeg incorporates the existing GazeFollow dataset and introduces a novel annotation protocol to improve data quality. The GazeSeg dataset has a wide range of application scenarios including social interaction, autism diagnosis and human-computer interaction, with the goal of enhancing the performance of gaze following tasks through precise pixel-level prediction.




