Optimized Sequential model for Plant Recognition in Keras

Abstract

Abstract There are huge varieties of floras in the world. Lots of varieties of species are beneficial to the human’s life. Plant recognition is a very important task to segregate the huge amount of floras which belong to various categories. Various researchers have applied different approaches to recognize the plant family. Deep learning is a subset of machine learning. This is one of the accepted technologies that automatically extracts features, processes them and yields the best results. Keras is a widely used deep learning framework which is employed in this work. Five different plant species are chosen as samples among the Indian species, namely OcimumTenuiflorum, Sansevieriatrifasciata, Chlorophytumcomosum, Azadirachtaindica, Aloe Vera. These samples are rich in oxygen. From these samples the features like shape, color, texture, corners are extracted. One hot encoding is also applied onto the target values to optimize the results of recognition. The extracted features are fed into the sequential keras model which recognizes the plant species. The accuracy of the training set is 100 percent and the testing set is 96.7percent. Confusion matrix is drawn to show the correctly classified and misclassified samples.