Machine Learning Based Flower Recognition System

Automatic flower plucking systems for smart agriculture are being studied for many years to support flower harvesting. Such systems require flower recognition task to be integrated as part of the system. This paper presents an approach for classification of flowers using a machine learning algorithm. The method categorizes flowers into different species with the help of convolutional neural networks and deep learning techniques. The system uses a pre-trained CNN model to improve the accuracy rate. Concepts such as Feedforward, back-propagation and transfer learning are used to create the neural network model. Different hyper-parameter values have been tested on the model which provides maximum accuracy of 85.0 percentage on the testing dataset. The result is visualized in the form of bar-plots which provides the top 5 predictions of flower species for the given input image of a flower.