Medical Leaf Classification using Deep Learning

This study suggests classifying medicinal plant leaves using a machine learning-based method. The following six types of medicinal plant leaves are included in a dataset that was gathered from the Department of Agriculture: peppermint (Mentha balsamea), bael (Aegle marmelos), lemon balm (Melissa officinalis), catnip (Nepeta cataria), stevia (Stevia rebaudiana), and tulsi (Ocimum sanctum). These leaves are captured digitally and in multispectral utilizing a computer vision lab configuration. Leaf sections are clipped and turned into gray-level pictures during the preprocessing stage. The Sobel filter is then used for edge and line detection based on seed intensity, and five regions of interest are drawn. Extracted are 65 fused features that combine multispectral properties, run- length matrix, and texture. The dataset is reduced to 14 optimized features by the use of the chi-square feature selection approach for feature optimization. The improved dataset is then subjected to five machine learning classifiers: bagging, random forest, simple logistic, multi-layer perceptron, and logit-boost. Of the classifiers studied, the multi-layer perceptron classifier shows the most potential accuracy.