论文标题
动态环境中的对象实例标识
Object Instance Identification in Dynamic Environments
论文作者
论文摘要
我们研究了人们与对象互动的动态环境中识别对象实例的问题。在这样的环境中，对象的外观通过与其他实体的相互作用，手动阻塞，背景变化等动态变化。这会导致外观内部范围更大的外观变化，而不是在静态环境中。为了发现这种情况下的挑战，我们新建立了在Epic-Kitchens数据集中构建的1,500多个实例的基准，该数据集包括自然活动并对其进行了广泛的分析。实验结果表明（i）针对特定实例的外观变化的鲁棒性（II）集成低级（例如，颜色，纹理）和高级（例如对象类别）特征（III）在重叠对象上的前景特征选择是进一步改进所必需的。
We study the problem of identifying object instances in a dynamic environment where people interact with the objects. In such an environment, objects' appearance changes dynamically by interaction with other entities, occlusion by hands, background change, etc. This leads to a larger intra-instance variation of appearance than in static environments. To discover the challenges in this setting, we newly built a benchmark of more than 1,500 instances built on the EPIC-KITCHENS dataset which includes natural activities and conducted an extensive analysis of it. Experimental results suggest that (i) robustness against instance-specific appearance change (ii) integration of low-level (e.g., color, texture) and high-level (e.g., object category) features (iii) foreground feature selection on overlapping objects are required for further improvement.
