Abstract
Black gram (Vigna mungo) is considered one of the most important pulse crops cultivated in Bangladesh because it is a vital source of nutrition and a potential source for raising a good income. It is one of those plants where most leaves are affected by diseases. We observed that most of the leaves were diseased in the fields, and we had difficulty collecting healthy samples. The crop is affected by different diseases attacking leaf tissues, causing heavy yield loss. We can apply deep learning models to recognize diseases in their early stages for timely interference. Diseases could be detected with the automation process, from which much enhancement in the management and yield of black gram crops is possible. Our purpose is to create a unique dataset of Bangladesh's Black Gram (Vigna mungo) to help global researchers build a deep learning-automated system for the early detection and classification of Black Gram leaf diseases that will assist farmers and create more awareness among different agricultural stakeholders. The original dataset of 4,038 images was collected from the Sirajganj and Solonga regions in Bangladesh. The dataset has five different classes: Healthy, Cercospora Leaf Spot, Insect, Leaf Crinkle, and Yellow Mosaic. This dataset will help researchers improve disease detection in Black Grams by developing effective computational models and applying advanced machine learning techniques.