Plant Leaf Disease Prediction Using Transfer Learning

Plant Leaf Disease Prediction Using Transfer Learning

HarisAhmed I Nidagundi, Mohammed Ismail I, Zafeer Ahmed, Sneha Sujit Saha, Prof. Sreenath MV

Department of Computer Science and Engineering
PES University
Bengaluru, India

Abstract—According to a poll conducted in 2018, agriculture employed around 80% of the labour force in India and generated between 16 and 18% of the GDP. It is regarded as the foundation of the Indian economy. A lot of work and research is being done to automate agricultural tasks, and with the development of technologies like Deep Learning, one of those tasks is the prediction of leaf disease. Identification of leaf disease is crucial for preventing the loss in the yield and quantity of the agricultural product. Machine learning and Image Processing techniques are used for detection of plant diseases. The strategies that already exist have limitations when it comes to the ability to adapt and reuse the results of learning. As a result, we employ the transfer learning approach to enhance performance and shorten the training period. We collected PlantVillage dataset from Kaggle. Images of 15 different plant leaf classes from the three different plants potato, pepper, and tomato are shown. The dataset was split into three separate datasets. The model provides a maximum accuracy of 99%. The final training accuracy was 87% and the final validation accuracy was 99%, with low training and validation loss. The experimental results showed that our model achieved a good accuracy rate for plant leaf disease prediction.

Keywords—Plant Disease, Machine Learning, Transfer Learning, Kaggle, Python.