Apple MOTS Dataset

The Apple MOTS: Multiple Object Tracking and Segmentation in an Apple Orchard Field dataset comprises temporally synchronized images and labels of apples, captured using UAVs (Unmanned Aerial Vehicles) and a wearable sensor in an orchard. It includes a total of 86,000 manually annotated apple instances and 1,700 frames annotated following the MOTS (Multi-object Tracking and Segmentation) style. Sequences 0 to 5 are allocated for training, while sequences 6 to 8 are designated for testing and validation. Sequences 10 to 12 represent the testing datasets, which feature overlays for "ignore regions."

苹果多目标跟踪与分割数据集（Apple MOTS）囊括了在果园中利用无人机（Unmanned Aerial Vehicles）及可穿戴传感器捕获的苹果的时序同步图像及标签。该数据集包含共计86,000个手动标注的苹果实例以及1,700帧遵循多目标跟踪与分割（MOTS）风格的标注图像。序列0至5被分配用于训练，而序列6至8则被指定用于测试与验证。序列10至12构成了测试数据集，其中包含“忽略区域”的叠加标注。

搜集汇总

背景与挑战

背景概述

Apple MOTS Dataset是一个专注于苹果园场景的多目标跟踪与分割数据集，包含2198张图像和105559个苹果实例标注，适用于实例分割、语义分割和物体检测任务。数据集由瓦赫宁根大学与研究机构于2022年发布，采用CC BY 4.0许可，分为训练集和测试集，并包含12个场景。