DGL-MOTS

DGL-MOTS数据集是由佛罗里达大学等机构创建，专注于自动驾驶中的多对象跟踪与分割任务。该数据集包含68,000个实例，覆盖多种驾驶场景，如高速公路、本地道路、停车场和住宅区，旨在提高网络训练的准确性和效率。数据集通过DG-Labeler工具进行标注，该工具利用深度粒度模块描绘实例空间关系，生成精细实例掩码。DGL-MOTS数据集在数据多样性、标注质量和时间表示方面超越了现有数据集，适用于提升未来交通视觉感知能力。

DGL-MOTS dataset was developed by institutions including the University of Florida, focusing on the multi-object tracking and segmentation task in autonomous driving. This dataset contains 68,000 instances, covering diverse driving scenarios such as highways, local roads, parking lots and residential areas, aiming to improve the accuracy and efficiency of network training. The dataset is annotated using the DG-Labeler tool, which leverages a deep-grained module to depict the spatial relationships among instances and generate fine-grained instance masks. The DGL-MOTS dataset outperforms existing datasets in terms of data diversity, annotation quality and temporal representation, and is applicable to enhancing future traffic visual perception capabilities.