VegNet: Dataset of vegetable quality images for machine learning applications

The agricultural industry has an unmet requirement for quick and accurate classification or recognition of vegetables according to the quality criteria. This open research problem draws attention to the research scholars every time. The classification and object detection challenges have seen highly encouraging outcomes from machine learning and deep learning techniques. The foundational condition for developing precise and reliable machine learning models for the real-time context is a neat and clean dataset. With this goal in mind, we have developed a picture dataset of four popular vegetables in India that are also highly exported worldwide. In order to generate a dataset, we have taken into account four vegetables: Bell Peppers, Tomatoes, Chili Peppers, and New Mexico Chiles. The dataset is divided into four vegetable folders, including Bell Pepper, Tomato, Chili Pepper, and New Mexico Chile. Further each vegetable folder contains five subfolders namely (1) Unripe, (2) Ripe, (3) Old, and (4) Dried (5) Damaged. The image collection includes a total of 6850 pictures of vegetables in dataset. We firmly feel that the provided dataset is very beneficial for developing, evaluating, and validating a machine learning model for vegetable categorization or reorganization.


Specifications
Agriculture Sciences, Horticulture, Vegetable Quality, Machine Learning Specific subject area Unripe, Ripe, Old, Dried and Damaged quality image dataset Type of data Vegetable images How data were acquired The high quality vegetable images were captured using mobile phone camera with different background and artificial light. Data format Raw Description of data collection The high resolution rear camera of mobile phone was used to capture the different stages of vegetables. The images were taken jpg. Format with the dimension of 4624 × 3472. The captured images then resized to 256 × 256 dimensions using python script. The resized image dataset is stored in four folders viz. Bell Pepper, Tomato, Chili Pepper, and New Mexico Chile. The vegetable images then segregated in five subfolders viz. Unripe, Ripe, Old, Dried and Damaged vegetable according to the vegetables quality. All the images were taken in different light condition with white background. This vegetable image dataset can be used in testing, training and validation of vegetable classification or reorganization model. Data  solutions to classify vegetables as per their quality. • The dataset can be used to create high-quality vegetable classification apps that are valuable for farmers, the agricultural sector, wholesalers, hawkers, and customers, as well as vegetable export businesses.

Data Description
As a fraction of all agricultural output, the vegetable market's profit share is sizable [1][2][3][4] . The greatest requirement in the agro-industry is for quick and accurate vegetable classification. Utilizing computer vision and deep learning techniques, the veggies may be divided into many groups based on their outward characteristics, such as shape, size, and color [5][6][7][8][9] . Vegetables with quality parameters for those that are heavily consumed or exported in accordance with Agricultural & Processed Food Products Export Development Authority (APEDA) are included in this VegNet dataset [10] . This dataset consists of four classes of vegetables namely Bell Pepper, Tomato, Chili Pepper, and New Mexico Chile. This dataset contains the images of these vegetables and not their plants' leaves. These vegetables are worldwide cultivated by traditional farming, plant tissue culture and hydroponics methods. It is mostly used in culinary and secondary metabolite production [11] . The main reason for choosing these 4 vegetables is the change in color with time. These vegetables contain the red coloured carotenoid 'lycopene' which causes the vegetables to change color when they ripen. This color changing ability will be effective in identifying the stages of vegetables; whether it is ripen, over ripen (old) or dried category.
In this dataset the images were captured using mobile phone and categorized into five subclasses namely Unripe, Ripe, Old, Dried and Damaged. Images of vegetables were captured on white backgrounds under various lighting conditions in both indoor and outdoor places. The VegNet dataset contains different folders which are created based on the vegetables quality and not on the image quality. Fig. 1 displays a various photos from dataset's, which were captured in a variety of settings.    Resolution unit 2 8

Materials or Specification of Image Acquisition System
Color representation sRGB

Method
All the four vegetable Bell Pepper, Tomato, Chili Pepper, and New Mexico Chile were purchased from local market in various stages. The vegetables brought to laboratory and washed it carefully (except dried and damaged). Daily photos were taken using a high definition rear camera of a Xioami Mi10T smartphone with various angles against white backdrops. The images were captured in a single as well as with multiple vegetables. The images were captured with different angle, color, background and lightning situation. Various photographs were captured in the original dimensions, which were 4624 × 3472. Using python script the images then converted to 256 × 256 dimension The created images are publicly available and uploaded online on Mendeley Data [13] . The classes, number of photographs taken, and environments where the images were taken are all listed in Table 4 .

Ethics Statement
The data is available in public. No ethics approval needed for this study.

Declaration of Competing Interest
The authors affirm that they have no known financial or interpersonal conflicts that would have appeared to have an impact on the research presented in this study.