An optimized hyper parameter-based CNN approach for predicting medicinal or non- medicinal leaves. (October 2022)

• Record Type: Journal Article
• Authors: Amulya, K.; K Deepika, Dr.; P Kamakshi, Dr.
• Is Part Of: Advances in engineering software. Volume 172(2022)
• Publication Date: 2022-10
• Subjects: Medicinal leaves -- CNN -- Features -- OHP- based CNN -- Gabor filter; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2022.103181
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: Medicinal leaves and non-medicinal leaves of about 2500 images were considered to detect rare species of medicinal leaves effectively. The primary focus is to extract features by applying the Gabor filter and other feature extraction techniques, which include resizing and cropping images. The feature extraction is the major part of the implementation and this is done by using Gabor filter. This model with the Gabor filter provides the best accuracy of 97%. Based on the trained model the results and performance metrics are evaluated to examine the model performance. Abstract: Medicinal leaves and Non-Medicinal leaves of about 2500 images were considered to detect rare species of Medicinal leaves effectively. The primary focus is to extract features by applying the Gabor filter and other feature extraction techniques, which include resizing and cropping images. The feature extraction is the major part of the implementation and this is done by using Gabor Filter. The OHP- Based CNN model (optimized hyper parameter-based CNN model) is applied to identify the best parameter to train the model based on the features extracted and then the data fed to the model is classified as medicinal and non-medicinal purposes This process of training the model helps to extract most of the features and improves the accuracy. This model with the Gabor filter provides the best accuracy of 97%. Based on the trained model the results and performance metrics are evaluated to examine the model performance. As a result, this approach aids to classify the medicinal leaves and non-medicinal leaves as well as in the identification of rare species of medicinal leaves and use them for the medicinal benefit and research purposes.