MedChecker- An Ensembled Deep-Learning-based Classification Model

Kerala is rich in its wide variety of flora and fauna and most of them have considerable medicinal values. The majority of the time, it is exceedingly challenging to manually identify the many species of these plants. In this study we have proposed an ensemble deep- learning model for the classification of medicinal plants available in Kerala. The final prediction of this model will be from two deep-learning based classification models. We have included only 14 classes of plants in this study. Plants having resemblance in their leaves have been categorized and trained using multiple models and they got ensembled. We aim to develop a free and open-source mobile application for presenting this work to the end-users. The front-end has been developed in Flutter and the classification models are built using the pre-trained model ResNet. The first model includes 10 classes and the second model is trained with varieties of Tulsi. The first model showed accuracy of 0.9766 and loss of 0.0777 during validation and 0.9856 and 0.0339 during training. Whilst, the accuracy and loss of the second model are 0.9241 and 0.0812 during validation and 0.9458 and 0.0314 during training.