Home About Login Current Archives Announcements Editorial Board
Submit Now For Authors Call for Submissions Statistics Contact
Home > Archives > Volume 20, No 7 (2022) > Article

DOI: 10.14704/nq.2022.20.7.NQ33348

Recognition of early blight and late blight diseases on potato leaves based on CNN algorithm

Dr Ashok Kumar Koshariya, Dr. B. Priyadharshini, Mrs. V. Ramalakshmi, Dr. Kalyani Pradhan, Mr. Jaskaran singh, S. Balamuralitharan

Abstract

Potatoes are a well-known vegetable crop grown all round the world. The crop, however, may additionally be harmed by way of way of potential of a complete lot of illnesses. It is quintessential for the planter to apprehend the form of contamination in order that the crop can be dealt with promptly. The leaves had been decided to be a easy sign of numerous diseases. For the detection of tomato crop illnesses, lots of Machine Learning (ML) algorithms, in addition to Convolution Neural Network (CNN) fashions, are superior withinside the literature. CNN fashions rely upon Deep Learning, Neural Networks, which can be excellent from regular Machine Learning techniques consisting of k-NN, Decision-Trees, and so on. While pre-skilled CNN fashions art work admirably, the sizeable range of matters involved makes them computationally intensive. A less elaborate CNN model with eight hidden layers is proposed in this painting. The suggested lightweight model outperforms the regular machine getting to be aware of techniques and pre-skilled fashions at the publically reachable dataset PlantVillage, with an accuracy of ninety-eight. 4 percent. The PlantVillage dataset contains 39 classes of numerous crops, consisting of apple, potato, maize, grapes, and so on, of which 10 are tomato illnesses. While k-NN has the very first-class accuracy of ninety four 9 percentage in classical ML methods, VGG16 has an exceptional accuracy of ninety three 5 percentage in pretrained fashions. Afterimage augmentation, picture pre-processing became into employed to decorate the general overall performance of the proposed CNN with the aid of way of potential of adjusting the brightness of the picture by way of skill of an possibility variable of a randomized width of the photograph. With an accuracy of ninety-eight percent, the recommended model moreover works enormously proper with datasets aside from PlantVillage.

Keywords

Convolution Neural Network (CNN), Image pre-processing and Machine Learning,

Full Text

PDF

References

?>