PaddyVarietyBD: Classifying paddy variations of Bangladesh with a novel image dataset

Abstract

Among countless crop varieties produced worldwide, the staple food of most of Asia, some parts of Europe, and North America is rice. Being an essential food item, rice offers an integral contribution to the economy of countries like China, India, Bangladesh, Pakistan, Indonesia, and so on. Scientists have long been working on developing new and improved rice species to battle different environmental hindrances and natural calamities. Although numerous research and studies have been conducted on this diverse crop, artificial intelligence, in particular, machine learning has not been practiced in this field with its full potential. The key factors behind this lag include the unavailability of standard and ready-to-use datasets. Intending to mitigate this drawback, this paper proposes an image dataset of paddy species to assist researchers and scientists in classifying, analyzing, and evaluating paddy classes. To the best of our knowledge, this is the first standard and open dataset of paddy varieties in Bangladesh. The rice sample was collected from two places namely – Bangladesh Institute of Nuclear Agriculture (BINA) and the Bangladesh Institute of Rice Research Institute (BRRI) where agrarian scientists work on developing new or improving existing paddy species. The dataset contains 14,000 RGB microscopic images of each paddy kernel. The enormity and inclusivity of the dataset make it useful for global research purposes. The dataset can be a useful resource not only in the area of artificial intelligence, but also in agriculture, botanical, and economic research.