Plant Leaf Disease Detection using Transfer Learning and Explainable AI Md Humaion Kabir Mehedi , A.K.M. Salman Hosain , Shafi Ahmed , Samanta Tabassum Promita , Rabeya Khatun Muna , Mehedi Hasan , and Md Tanzim Reza Department of Computer Science and Engineering Brac University 66 Mohakhali, Dhaka - 1212, Bangladesh {humaion.kabir.mehedi, akm.salman.hosain, shafi.ahmed, samanta.tabassum.promita, rabeya.khatun.muna, mehedi.hasan1}@g.bracu.ac.bd rezatanzim@gmail.com Abstract—Among the major occupational sectors around the world, agriculture has the highest level of involvement. Every year, this sector faces a substantial loss in production and profit due to a large number of diseases in crops and plants. If those diseases are not detected early and taken measures for prevention, it can bring about a devastating result that can financially burden agriculture personnel. Traditional methods of detecting diseases in plants and crops offer high accuracy. However, the procedure is time-consuming, which might be insidious in most cases. Crop diseases need to be detected and cured as soon as possible as most diseases are highly contagious among crops and plants. In this paper, we have used the transfer learning approach with three pre-trained models: EfficientNetV2L, MobileNetV2, and ResNet152V2, to detect various plant diseases. We have proposed a framework to detect 38 types of leaf diseases in 14 different plants, compared the three pre-trained models based on various quantitative performance evaluation parameters, and demonstrated that EfficientNetV2L performed best with 99.63% accuracy. In the end, Explainable Artificial Intelligence (XAI) technique: LIME has been implemented in our model to un- derstand the insight view of the model EfficeintNetV2L’s for such prediction. It is used to make our model’s predictions more reliable and gives a clear explanation about the reason of such decision. Index Terms—EfficientNetV2L, ResNet152V2, MobileNetV2, Transfer Learning, Plant Leaf Disease, XAI, LIME I. I NTRODUCTION In 2018, around 28% of the global population worked in agriculture, feeding 1 billion people. In 2018, around 4% of global GDP was from agricultural sector [1] [2]. However, due to population expansion, food demand rises. To meet this need, increasing agricultural productivity and protecting crops are of vital importance [3]. Crop diseases are a severe problem in agriculture since they impact both the quality and volume of agricultural production [4]. Due to the presence of several pathogens in their environment, crops are very susceptible to various diseases [3]. Crop disease can affect production from 10% to 95% [5]. Some disease-causing agents are infectious viruses, while others are fungi, bacteria, or parasites. Infected plants or plant detritus can infect healthy plants [6]. The annual production loss of crops as a result of various plant diseases accounts for 10 to 40 percent for five of the world’s most crucial crops [7]. Plant disease threatens food security. Detecting plant diseases prematurely is essential to increasing the food supply. Advanced crop disease diagnosis and prevention are required to minimize disease-induced crop damage and maximize agricultural output [8]. Pest control is as old as agriculture because of the ancient link between disease and future harvest. Early farmers used supernatural or superstitious tactics to combat crop pathogens. Modern plant pathology enables the scientific diagnosis of plant diseases [9]. Direct visual identification of plant disease symptoms on leaves or chemical procedures using molecular studies on leaves are traditional methods [9]. However, these procedures are slow and labour-intensive [9].Most farmers in impoverished nations and small farms identify crop illnesses visually [6]. As a result, they often end up diagnosing the wrong disease and using medicines or pesticides based on the wrong diagnosis. It exacerbates the condition rather than ameliorates it. If a field is contaminated with a rare illness, farmers seek specialist help for a correct diagnosis, which increases treatment costs [6]. The motivation of our work was to help agricultural per- sonnel to avoid these insidious circumstances by detecting plant diseases in a faster and more accurate way. Nowadays, machine learning algorithms are used in various sectors. A timely assessment of the problem is necessary to avert significant damage and enhance production. In recent years, promising ways for detecting and localizing diseases utilizing automatic monitoring and recognition systems have been de- veloped [4].Machine learning and deep learning approaches can detect plant illnesses fast. In this paper, we have utilized a transfer learning approach with explainable AI to tackle the challenge of early detection of plant diseases. Our contribu- tions to this paper are- • We have used three pre-trained models, EfficientNetV2L, 978-1-6654-6316-4/22/$31.00 © 2022 IEEE