Search

Reliable and accurate detection of fruits during the whole growth period has always been one difficult and important bottleneck for achieving precise, intelligent and efficient orchard management. In order to deal with the insufficiency of sample scale and diversity in actual production scenes, this dataset is constructed focusing on the application of fruit detection in typical orchard operation stages, such as fruit thinning, bagging and picking operations. Specifically, through in-field shooting and data post-processing, the multi-modal image dataset for peach detection is released, which covers the acquisition, classification, labeling, storage and use of multi-modal peach images during fruit thinning, bagging and picking stages under the different natural circumstances, including complex weather, illumination and occlusion. What’s more, the modalities of the dataset get involved in the types of visible light, depth and infrared with the total storage amount of 8.27GB, providing fundamental and valuable image resources for the following research areas, e.g., multi-modal image data fusion and object detection. In addition, the dataset can also be used as a standard library for deep learning modeling in big data environment with the important practical application value for promoting the research of fruit object detection., 果实目标全生长期的可靠精准检测一直是实现果园精准智能高效生产管理亟待突破的难点和重要瓶颈。针对当前毛桃果实目标识别领域数据多样性不足、实际生产场景样本匮乏的情况，本数据集面向毛桃疏果、套袋、采摘等关键果园作业阶段的果实目标检测应用，通过实地拍摄和数据处理，构建了毛桃多模态目标检测数据集，涵盖疏果期、套袋期和采摘期高质量毛桃多模态图像的采集、分类、标注、存储与使用等多方面的内容。其中，毛桃多模态图像数据集获取场景涵盖了多种天气、复杂光照和多类遮挡情况，数据类型包括可见光、深度和红外多模态图像数据，共计8.27GB。本数据集将为多模态图像数据融合和目标检测等研究方向提供宝贵的基础图像数据资源，同时可作为大数据环境下+深度学习建模的标准图库，对促进果实目标检测领域研究具有重要的实际应用价值。