Abstract
Vegetables are one of the most essential parts of the agricultural sector and the food supply chain; therefore, the identification and categorization of vegetable types require effective strategies. In this paper, we introduce the Vegetable Image Dataset, which is a meticulously developed collection of 4319 images representing 12 different vegetable species native to Bangladesh, including Potato, Onion, Green Chili, Garlic, Radish, Bean, Ladies Finger, Cucumber, Bitter Melon, Brinjal (Eggplant), Tomato, Pointed Gourd. The dataset contains images taken in natural environments, including local markets, agricultural fields, and homes, using phone cameras to represent real-world conditions better. All photos have undergone background removal and annotation to highlight features such as shape, texture, and color, thus making it a handy resource for deep-learning projects. Developed primarily for developing convolutional neural network (CNN) models, this dataset allows for the automatic identification and classification of vegetables for various applications. Applications range from improving the supply chain for agriculture to allowing instantaneous detection of vegetables in kitchens or marketplaces and increasing the efficiency of automation for sorting and packaging. With its unique characteristic of Bangladeshi vegetables, this dataset provides the valuable resource needed for improving agricultural practices using AI-driven ways and fostering further developments of technologies in underserved communities.