Abstract:
Plant identification is an interdisciplinary focus of both botanical taxonomy and computer vision. Plant leaves have an assortment of features like design, shape, size, texture, color, and venation. These features are studied by botanists in order to identify the plant species. Today's botanists have the advantage of using computer vision based technology for automated leaf analysis and identification. Currently, computer vision researchers are increasingly inclined towards Convolutional Neural Networks and Deep Learning methods for feature extraction and classification problems given its efficiency and accuracy. This paper presents a Deep learning framework for leaf classification developed by transfer learning Resnet pretrained model. Mere transfer learning of Resnet was found inadequate for accurate classification of the heterogeneous dataset. Hence augmentation was performed to increase the samples in the dataset for a more powerful classification. The proposed methodology is evaluated on the Malayakew leaf dataset and was found to be efficient.
|