The lemon dataset can be used to study the possibility of solving the fruit quality control problem. It contains 2,690 annotated images (1056 x 1056 pixels). The original lemon images were obtained in 2020 using the blog post “When life gives you lemons create a dataset!, and using the process described in CVAT Manual annotation.

The file name consists of specific identifiers, such as 0037_G_I_120_A:(0037 Single fruit instance), 120(Relative to the photo angle), A(Photo location).

Can be used COCO API to read the dataset.

from pycocotools.coco import COCO coco = COCO('../lemon-dataset/annotations/instances_default.json')