ZeroWaste
收藏资源简介:
ZeroWaste数据集是由波士顿大学等机构联合创建的,专注于工业级废物检测与分割的首个野外数据集。该数据集包含4661帧高密度标注的图像,用于训练和评估检测与分割模型,以及大量未标注图像用于半监督和自监督学习方法。数据集内容涵盖高度变形和半透明物体,以及物体类别间的细微差异,为自动化视觉识别带来独特挑战。ZeroWaste数据集的应用领域包括提高回收效率、增加利润和保障工人安全,旨在解决废物分类过程中的低效问题。
The ZeroWaste dataset, co-developed by Boston University and other institutions, is the first real-world wild dataset dedicated to industrial-scale waste detection and segmentation. It contains 4,661 densely annotated image frames for training and evaluating detection and segmentation models, along with a large number of unlabeled images designed for semi-supervised and self-supervised learning methods. The dataset covers highly deformed and translucent objects, as well as subtle differences between object categories, presenting unique challenges for automated visual recognition. The application scenarios of the ZeroWaste dataset include improving recycling efficiency, boosting profits, and ensuring worker safety, with the core objective of addressing inefficiencies in the waste sorting process.




