Object Placement Assessment (OPA) dataset

OPA数据集由上海交通大学和北京理工大学创建，专注于评估合成图像中物体放置的合理性。该数据集包含73,470张合成图像，每张图像都有关于物体放置合理性的二元标签。数据集的构建过程包括从COCO数据集中选择未遮挡的物体和背景图像，随机调整物体的大小和位置，并由人工标注者进行合理性标注。OPA数据集的应用领域包括增强现实和艺术创作，旨在解决物体在合成图像中放置不合理的问题。

The OPA dataset was created by Shanghai Jiao Tong University and Beijing Institute of Technology, focusing on evaluating the plausibility of object placement in synthetic images. This dataset contains 73,470 synthetic images, each paired with a binary label indicating the plausibility of the object's placement. The construction process of the OPA dataset includes selecting unoccluded objects and background images from the COCO dataset, randomly adjusting the size and position of the objects, and having human annotators perform plausibility annotation. The application fields of the OPA dataset cover augmented reality and art creation, with the goal of addressing the issue of implausible object placement in synthetic images.