QuinceSet: Dataset of annotated Japanese quince images for object detection

• DOI: 10.1016/j.dib.2022.108332
• Publication Date: 2022-05-29T13:54:38Z
• Authors: Edīte Kaufmane; Kaspars Sudars; Ivars Namatēvs; Ieva Kalniņa; Jānis Judvaitis; Rihards Balašs; Sarmīte Strautiņa
• Keywords: Ground truth

With long-term changes in temperature and weather patterns, ecologically adaptable fruit varieties are becoming increasingly important agriculture. For selection of candidate cultivars breeding or for yield predictions, set characteristics at different growth stages need to be described evaluated, which is largely done visually. This a time-consuming labor-intensive process that also requires sufficient expert knowledge. The annotated dataset Japanese quince - QuinceSet consists images (Chaenomeles japonica) fruits taken two phenological developmental detection phenotyping. First, after flowering, when the second fall over have reached 30-50% their final size, second, ripening stage quince, just before yielded. Both classified as unripe ripe were using ground truth ROI presented YOLO format. contains 1515 high-resolution RGB .jpg with same number .txt files. Images manually LabelImg software. A total 17,171 annotations provided by experts. acquired on site Institute Horticulture Dobele, Latvia. Homogenization was performed under conditions, times day, from capturing angles. both fully visible quinces partially obscured leaves. Care ensure foreground, leaves has adequate brightness minimal shadows, while background darker. will allow increase efficiency estimation, identify phenotype more reliably, may useful other crops.