Browsing by Author "Gunasinghe, H.N."

Now showing 1 - 2 of 2

Gender recognition of Luffa flowers using machine learning
(International Research Conference on Smart Computing and Systems Engineering - SCSE 2018, 2018) Gunasinghe, H.N.; de Silva, R.
Automatic flower gender identification could be introduced to large farmlands to help artificial pollination of imperfect flowers. Incomplete flowers contain either male or female organs but not both. In this paper, we present a computer aided system based on image processing and machine learning to identify the gender of a Luffa flower automatically. A pre-trained machine learning model is used for gender segmentation of flowers. The system is developed using Tensorflow Machine Learning Tool, which is an open-source software library for Machine Intelligence. The network was selected as the Google’s Inception model and a dataset was prepared after capturing flower images from a Sri Lankan Luffa farm. The system was tested using two datasets. The first contained the captured original images and the second was prepared by cropping each image to extract male and female floral organs, stamen and pistil respectively. The prototype system classified the flowers as either male or female at 95% accuracy level. The experimental results indicate that the proposed approach can significantly support an accurate identification of the gender of a Luffa flower with some computational effort.
Identification of Papaya Fruit Diseases using Deep Learning Approach
(4th International Conference on Advances in Computing and Technology (ICACT ‒ 2019), Faculty of Computing and Technology, University of Kelaniya, Sri Lanka, 2019) Munasingha, L.V.; Gunasinghe, H.N.; Dhanapala, W. W. G. D. S.
The diseases are a major problem faced by all the farmers including fruit farmers. It is a threat for large farmlands because these diseases spread throughout the land and make the fruits inedible, which at the end impact badly on the farmer’s income. Hence early disease detection is very important for the farmers to prevent or to control the propagation of the diseases. The traditional method of fruit disease detection and identification is naked eye observation. Even if this method is sufficient for a home gardener, it is a very inefficient one that requires experience and expertise. As a solution for this problem several computerized approaches are being developed using Machine Learning and Image Processing techniques in the resent researches. In our proposed work, we considered Papaya fruit, as it is a very popular fruit cultivation in Sri Lanka. In this study we have implemented a computerized model for papaya disease identification using Convolutional Neural Network (CNN). Among various diseases of papaya fruit, anthracnose, black spot, powdery mildew, phytophthora and ringspot were chosen. These are commonly found in Sri Lankan papaya cultivation. We have collected diseased images using a digital camera in normal conditions from papaya farms. Some of the images were found from the publicly available images on the internet. Then we have trained a deep CNN for these images. The network is able to classify images into five major papaya diseases. The system can finally identify the disease once a new image fed to it. The model performed ~92% of classification accuracy for new images. With compared to previous research done using Support Vector Machine (SVM), there is an increase of ~2%. Furthermore, it could be seen that the Black Spot disease was identified very easily by the model. Powdery Mildew was the most difficult disease to recognize. The results of this study reveal that this method is an accurate, reliable and efficient where it could be useful as an aid for expertise.