Validation of a Neural Network Based Leaf Classification Algorithm

C. S. Sumathi, A.V. Senthil Kumar


Plants are living organism which belongs to vegetable kingdom that can live both on land and water. More than 300,000 species of
plants exists on earth. In order to effectively conserve and save the genetic resources, plants are to be identified and leaf shape plays a significant
role in plant classification. In this paper, since identifying the relevant feature is of vital importance the features are extracted using information
gain based feature selection method. A feed forward neural networks with different learning methods viz., Levenberg-Marquardt learning,
Incremental Backpropagation learning and Batch Back propogation learning automate the leaf recongnition for plant classification. Comparison
shows that information gain helps select features that show good improvement on feed forward neural network (normalized cubic spline) with
batch back propogation classifier algorithm out performs with an accuracy of 95.56%.


Key words: Information gain, Levenberg-Marquardt, Incremental Backpropagation, Batch Back propogation, Feed Forward Neural Network

Full Text:




  • There are currently no refbacks.

Copyright (c) 2016 International Journal of Advanced Research in Computer Science