Open Access
Research Article
Hyperparameter tuning of CNN based potato plant disease detection model integrated with GridSearchCV
*Kajal KaulCorresponding authorkajalkaulphd@gmail.comUniversity School of Information, Communication and Technology Guru Gobind Singh Indraprastha UniversityDwarka, New Delhi, 110078, IndiaView full profile → , Anuradha Chuganuradha@ipu.ac.inUniversity School of Information, Communication and Technology Guru Gobind Singh Indraprastha UniversityDwarka, New Delhi, 110078, IndiaView full profile → , Amit Prakash Singhamit@ipu.ac.inUniversity School of Information, Communication and Technology Guru Gobind Singh Indraprastha UniversityDwarka, New Delhi, 110078, IndiaView full profile → , Ravinder Kumarravinder.kumar2@icar.gov.inDivision of Plant Pathology Indian Council of Agricultural Research - Indian Agricultural Research Institute Pusa, New Delhi, 110008, IndiaView full profile → , V. Shanmugamshanpatho@yahoo.comDivision of Plant Pathology Indian Council of Agricultural Research - Indian Agricultural Research Institute Pusa, New Delhi, 110008, IndiaView full profile →
* Corresponding author · click or hover a name for details
- Received:
- 07 Aug 2024
- Published Online:
- 01 Jan 2025
- Article type:
- Research Article
- Language:
- EN
- Article no.:
- JIOS-1846
- Pages:
- 1–14
Abstract
There is an increasing need for enhancing plant health in the potato crop since Potato is the staple diet for majority of population living in Asia. Potato crop yield is of prime importance for farmers and agriculturalist. The Deep Convolution Neural Network model has been optimized on the PV dataset for classification, detection and prediction of various Potato Plant Diseases like PEB, PLB and distinguishing it from healthy potato leaves which served as the class labels. Images were then preprocessed using various Data augmentation techniques for enhancing the computational efficacy of the models. This paper aims at comparing various hyperparameter tuned cases for hyperparameters namely optimizer, learning rate, epochs, keeping batch_size, filter_size, activation function and the number of channels fixed for CNN as the base model integrated with GridSearchCV technique. Amongst all the hyperparameters, it was found that optimizer Adam along with 0.0001 learning rate, 3*3 filter size, 32 batch size, 100 epochs and relu activation function at dense layer and softmax function at fully connected layer performed the most optimal. The CNN base model was trained and tested to predict the output class. The efficiency of the model was measured by accuracies achieved for the model including train_accuracy, validation accuracy and test accuracy as 0.959, 0.960 and 0.988 respectively. The losses for the model such as train_loss, validation loss and test loss were 0.102, 0.100 and 0.039 respectively. After comparing all the model cases, Precision was found out to be 0.988, F1 score was found to be 0.988, Recall was 0.988, Roc-auc value was 0.563. To further comprehend the performance and select the most optimal model GridSearchCV technique was then applied for measuring performance with regards to achieving Generalization, Regularization and Optimization in the model. With a high cross-validation score of 0.78 indicates that the model can function well on unknown data, regularization as indicated by high value of C i.e. 31.622, is more tolerant of misclassifications, potentially leading to higher accuracy but potentially also lower sensitivity to detecting positive cases and optimization as the model’s parameters were successfully optimized by the ‘newton-cg’ solver, resulting in strong performance.
Keywords
Subject Classifications
68T0568T07
References
[1] N. Sareen, Anuradha Chug, and Amit Prakash Singh, “An image based prediction system for early blight disease in tomato plants using deep learning algorithm,” Journal of Information and Optimization Sciences, vol. 43, no. 4, pp. 761-779 (2022).[2] M. Al-Adhaileh, A. Verma, T. Aldhyani, and D. Ko, “Potato Blight Detection Using Fine-Tuned CNN Architecture,” Mathematics, vol. 11, no. 6, pp. 1-16 (2023), doi: 10.3390/math11061516.[3] J. Chen, X. Deng, Y. Wen, Z. Liu, and A. Zhang, “Weakly-supervised learning method for the recognition of potato leaf diseases,” Artificial Intelligence Review, vol. 56, pp. 7985–8002 (2023).[4] S. Hassan and A. Maji, “Plant Disease Identification Using a Novel Convolutional Neural Network,” IEEE Access, pp. 1-13 (2022), doi: 10.1109/ACCESS.2022.3141371.[5] R. Sharma, A. Singh, K. Kavita, N. Jhanjhi, M. Masud, E. Jaha, and S. Sk, “Plant Disease Diagnosis and Image Classification Using Deep Learning,” Computers, Materials and Continua, vol. 71, pp. 2125-2140 (2022), doi: 10.32604/cmc.2022.020017. [6] D. Tiwari, M. Ashish, N. Gangwar, A. Sharma, and P. Kumar, “Potato leaf diseases detection using deep learning,” in Proc. 2020 4th Int. Conf. Intelligent Computing and Control Systems (ICICCS), Madurai, India, pp. 461–466 (2020), doi: 10.1109/ICICCS48265.2020.9121067.[7] N. Sinshaw and B. Assefa, “Artificial Intelligence Assisted Early Blight and Late Blight Potato Disease Detection Using Convolutional Neural Networks,” Ethiop. J. Crop Sci., vol. 8, no. 2, pp. 15-28 (2020).[8] G. Polder, P. M. Blok, H. A. C. de Villiers, J. M. van der Wolf, and J. Kamp, “Potato Virus Y Detection in Seed Potatoes Using Deep Learning on Hyperspectral Images,” Front Plant Sci., vol. 10, pp 1-13 (2019), doi: 10.3389/fpls.2019.00209, PMID: 30881366, PMCID: PMC6405642. [9] P. Mathur, F. Sheth, D. Goyal, and A. Gupta, “Deep insight: Mathematical modeling and statistical analysis for mango leaf disease classification using advanced deep learning models,” Journal of Interdisciplinary Mathematics, vol. 27, pp. 317-342 (2024), doi: 10.47974/JIM-1830.[10] D. Bhoyar, S. Jain, J. D. Kene, Y. A. Suryawanshi, A. R. Bhagat, and S. J. Karale, “Enhancing agriculture productivity with machine learning in crop yield optimization,” Journal of Statistics and Management Systems, vol. 27, no. 2, pp. 213–223 (2024), doi: 10.47974/JSMS-1248.
Views: 273Downloads: 10Citations: 1
