Gender identification of imperfect flowers using image classification

In this paper, we present two approaches that can be used to classify images of imperfect flowers according to their gender. By taking photos of small sections of the flower and that of objects in its neighborhood, it is possible to classify the imperfect flowers. Using images of bitter gourd flowers and various other objects, we tested our approaches. In the first approach, we used multilayer perceptron network in WEKA 3.8.0 on red, green and blue (RGB) components of the images and could classify all stamens and pistils correctly. In the second approach, we constructed a set of rules to classify images with hundred percent accuracy based on their RGB values. This has much less computational burden than the first approach.