An Approach to identify plant species using Keras ConvNets and Data Augmentation
Now-a-days, there are nearly millions of plant species available across the globe. Identifying a specific plant has been a major issue due to different properties and duplications. Convolutional Neural Networks (ConvNets) is a type of neural network specially designed for image classification and recognition.
The present paper is aimed at designing a novel technique which forms a ConvNet using data augmentation techniques and feeding pre extracted features into a multi layer perceptron to identify different variants of plant species. Keras functional API is used to build a combined model using tensor flow as back end that inputs images as well as pre extracted features of plant leaves. The proposed model reduces the overfitting (which is one of the issues faced by ConvNets) and improves the ability of convolutional Neural Networks.