Image dataset of important grape varieties in the commercial and consumer market

This work presents a primary dataset collected from various geographic locations in Iraq for the seedlings of eight varieties of grapes that are used for local consumption and export. Grape types included in the dataset are: deas al-annz, kamali, halawani, thompson seedless, aswud balad, riasi, frinsi, shdah. Leaves of each type of the seasoned fruit were photographed with high resolution device. A total of 8000 images (i.e., 1000 images per category) were captured using random sampling approach while maintaining the balance and diversity within grape image data. The proposed dataset is of significant potential impact and usefulness with features including (but not limited to) 8 varieties, that have different tastes and can support various industry in agriculture and food manufactures.


a b s t r a c t
This work presents a primary dataset collected from various geographic locations in Iraq for the seedlings of eight varieties of grapes that are used for local consumption and export. Grape types included in the dataset are: deas alannz, kamali, halawani, thompson seedless, aswud balad, riasi, frinsi, shdah. Leaves of each type of the seasoned fruit were photographed with high resolution device. A total of 80 0 0 images (i.e., 10 0 0 images per category) were captured using random sampling approach while maintaining the balance and diversity within grape image data. The proposed dataset is of significant potential impact and usefulness with features including (but not limited to) 8 varieties, that have different tastes and can support various industry in agriculture and food manufactures.

Value of the data
• This dataset can be proven useful for the identification of the types of grape seedlings mainly before their plantation and growth [ 1 , 2 ]. • The early identification of the desired grape type is critical due to the long seeding duration (approximately # of years) along with the effort and cost associated with their growth [3] . • Conventional methods for the identification of the type for a seedling requires expert knowledge while at the seeding site where the effort and cost associated with the involved labor can be significant [4] . • The development of a classification model for grape class identification with higher accuracy can be an important undertaking providing significant advantage with respect to the economic factors as well as timely identification. • The proposed dataset can be proven useful for computer science community, particularly computer vision, machine learning and deep learning to build robust grape classification models that could accurately classify grapes of various types.
• Models of such nature can be utilized by farming community before plantation to cut the time and cost associated with planting grapes of the desired types and reduce the risks of planting incorrect types. Additionally, the proposed dataset provides an opportunity to the research community to build machine or deep learning-based classification models for the detection of the plant health status (i.e., diseases free plants) for the included types of grapes.

Objective
To reduce the time, cost, and effort in identifying the desired seedlings of grape cultivars prior to planting, the proposed dataset is configured to build a model capable of classifying eight cultivars of grapes desired in Iraq.

Data Description
The dataset consists of eight folders (corresponding to the included grapes types), each folder contains 10 0 0 images of JPG format. The images within the dataset are all of the resolution of 60 0 0 × 40 0 0 pixels. Because of the high resolution of the images, the size of the images became 49.8 GB, which is a large size for uploading and downloading the dataset from the Internet, so the resolution of the images was changed using a tool resize pictures in Windows to be the dimensions 1620 × 1080 pixels, Thus, the data size is 1.95 GB, After compressing with a zip program, the data size became 1.83 GB.Each folder represents one of the eight grape varieties (Deas Al-Annz, Kamali, Halawani, Thompson Seedless, aswud balad , riasi ,frinsi, shdah). The grape leaves have been photographed as shown in Fig. 1 .

Experimental Design, Materials and Methods
Acquisition of the images for each grape type followed the workflow of based on a random leaf selection approach as illustrated in Fig. 2 . In the proposed approach, selection of the constituent of the population of each grape type followed the uniform distribution with an equal probability of being selected.
Data diversity was maintained via random leaf selection approach, by which both the leaves to be photographed and tree for each grape type were selected randomly.
In the post image acquisition process for each grape type, images per each grape type were transferred from the camera's memory to an external hard drive and kept in a folder bearing the name of the grape type. The acquisition of the images for the next grape type would then follow after the removal of the images for the transferred grape type, and so on. Leaves containing dust were not photographed at a high rate, as they represent noise, and the leaves that were eaten were not photographed. Photographed with the knowledge and consent of the owners of the gardens.

Ethics Statement
It should be noted that the work meets the ethical requirements.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests.