Automated Indigenous Plant Recognition and Medicinal Value Extraction System

In an age where environmental awareness and exploration of natural resources are paramount, identifying traditional and indigenous plant species has become crucial across diverse domains. This aim of this work is ultimately develop a mobile application framework that detects and recognizes indigenous medicinal plants of our country, its medicinal value and usage in indigenous food preparation. Moreover the key component essential is the right machine learning model which is scalable, incremental and more accurate. The proposed work is developing a robust and scalable machine learning model for detecting and recognizing indigenous medicinal plants species. Image data set pertaining to whole plant and parts like flower, leaf, fruit, stem, bark, root of unique 2626 Indian plant species were collected and used for training and testing three deep learning CNN models like Resnet50, MobileNetV2 and VGG16 for plant species identification. These three models performed with 85%, 82% and 87% accuracy individually. Finally stochastic weight averaging ensemble of the three independent approaches resulted an overall accuracy of 95%.