A Convolution Neural Network Based Approach for Detection of Tree from Leaf Images

Abstract

We are going through a time when global warming is a vital issue. No other living things but yes trees only can stand as opposition of global warming. Though computer aided plant recognition has gained much interest in recent years. Also this proved as a most important tool in such areas like pharmacological science, forestry and agriculture. There are 3.041 trillion plants all around the world. They are at least 60,065 different kinds. Many of the plants of rare kinds become extinct and rare plant identification through this investigation can be a contribution to save the existence of them. We have made a system that can identify those plants by their leaves. We take 5 usual leaves from our surroundings and our system can successfully identify those plants. Computer vision based plant recognition is a challenging problem due to the variable appearance of trees, high intra-class and small inter-class variability, complex geometry and multi-scale hierarchical structure. Our contribution through this project is identifying a plant by an image of a leave using computer vision technique deep learning. Extensions to related works are discussed including other object detection such as different kinds of bicycles (mountain bike, highway bike) detection.