A dataset of multimodal pear images for target detection

Accurate fruit detection is the premise of growth monitoring and intelligent orchard management. In the past decades, with the development of deep learning and multi-modal research technology, the anti-interference and generalization ability of the model to complex environment has been significantly improved, however it is in the demand of large-scale training samples. Therefore, a good image dataset for computer vision modeling is very crucial for efficiently implementing fruit target detection. Early fruit detection methods mainly depended on the color, texture and other differences of RGB images. It has been witnessed that it was difficult to completely describe the fruit target area in one accurate and effective way. With the introduction of depth, infrared images and other data, it was helpful to improve the fruit detection result by using the spatial relationship between scene and objects. Therefore, we have built a multi-modal image dataset for crisp pear target detection, which covers several aspects of crisp pears including multi-modal images acquisition, classification, labeling, storage and usage. A total of 5.86GB of high-quality multi-modal image data were collected to the multi-modal image dataset, under the different natural circumstances of cloudy and sunny, daytime and nighttime, lighting and backlighting, naked and bagging, motion blurring scenes. Different from the existing fruit datasets that only provide visible images, this image data set is composed of traditional RGB images, as well as corresponding depth images and infrared images, which can be used as training samples for modeling intelligent fruit multimodal vision. As a standard dataset for deep learning modeling in big data environment, this image dataset offers valuable basic data resource for the research directions of multi-modal image data fusion and target detection, with the important practical application potential of promoting the research in the field of fruit target detection.