Iranian date fruits ripeness classification and synthetic image generation

Published: 24 March 2025| Version 1 | DOI: 10.17632/jt56fcysvj.1
Contributors:
,
,
,
,

Description

Dataset of color images of Iranian dates of the Berhi variety. The images correspond to three different stages of ripeness of the dates: Khalal (unripe dates), Rotab (ripe) and Tamar (overripe). The challenge is to develop an accurate and fast method for classifying the images. Original images of date fruits were captured at a resolution of 883x556 pixels in Iranian date orchards. Prior to model ingestion, each image was pre-processed by cropping it to a standardized 256x256 pixel square. There are 50 images for each class, making a total of 150 images. The dataset also contains synthetic images. In this case, the challenge is to develop a method for generating realistic date images in the three stages. The synthetic images were generated using StyleGAN3, with horizontal flipping applied during training to increase data diversity. There are 480 synthetic images for each class, making a total of 1440 images. The images are organized in three folders: - Original. Contains the original images captured in 883x556 pixels in JPG format. - Preprocessed. Contains the images cropped to 256x256 pixels in JPG format. - Synthetic. Contains the synthetic images generated with StyleGAN3, with size 256x256 pixels in PNG format.

Files

Institutions

Sharif University of Technology, University of Mohaghegh Ardabili, Gorgan University of Agricultural Sciences and Natural Resources, Universidad de Murcia

Categories

Computer Vision, Image Processing, Fruit, Classification System, Food Ripening, Date Palm

Licence