Intra-AFruit
收藏资源简介:
Intra-AFruit是由墨尔本大学创建的一个大型合成数据集,专注于水果和蔬菜的类内遮挡场景。该数据集包含255,000张图像,每张图像都标注了多种类型的掩码、边界框、双重顺序关系以及完整的外观信息。数据集的创建过程利用了自动数据合成技术,确保了标注的准确性和详细性。Intra-AFruit不仅适用于类内实例分割任务,还支持其他与类内遮挡相关的任务,如遮挡对象的顺序感知和不可见部分纹理的推断。该数据集的应用领域包括自动化收获、机器人库存建设和水果质量控制,旨在解决类内遮挡场景下的视觉感知问题。
Intra-AFruit is a large-scale synthetic dataset developed by the University of Melbourne, focusing on intra-class occlusion scenarios involving fruits and vegetables. This dataset comprises 255,000 images, each annotated with multiple types of masks, bounding boxes, dual sequential relationships, and comprehensive appearance information. The dataset’s creation process leverages automated data synthesis techniques to ensure the accuracy and thoroughness of its annotations. Intra-AFruit is applicable not only to intra-class instance segmentation tasks, but also supports other occlusion-related tasks such as order perception of occluded objects and texture inference for invisible parts. Its application fields include automated harvesting, robotic inventory construction, and fruit quality control, aiming to address visual perception challenges in intra-class occlusion scenarios.




