Publication Type : Book Chapter
Publisher : Springer
Source : In Data Engineering and Intelligent Computing, pp. 101-108. Springer, Singapore, 2022.
Url : https://link.springer.com/chapter/10.1007/978-981-19-1559-8_11
Campus : Coimbatore
School : School of Engineering
Department : Center for Computational Engineering and Networking (CEN)
Year : 2022
Abstract : Unmanned aerial vehicles (UAVs) usually capture large amounts of images. The images need not be of good quality and need not contain the data that are required. Processing and selecting valuable data from these images take time. To overcome this difficulty, deep convolutional neural network algorithms can be assigned for processing huge image data. Deep convolutional neural networks (DCNNs) seem to provide substantial improvements in training efficiency and performance, primarily for object recognition tasks. The DRU-Net architecture used in this paper takes advantage of DenseNet and ResNet architecture to detect the target epiphyte in the images. The target epiphyte in this work is Werauhia Kupperiana which comes under the bromeliad family. The architecture was evaluated on a private dataset. The evaluation metrics used were IoU and the dice score. The results obtained through this study illustrate the capability of DRU-Net to detect the target epiphyte and cut down the botanist’s time in the segmentation of the target epiphyte and the background in the images.
Cite this Research Publication : Menon, Anjana K., V. V. Sajith Variyar, Ramesh Sivanpillai, V. Sowmya, Gregory K. Brown, and K. P. Soman. "Epiphyte Segmentation using DRU-Net." In Data Engineering and Intelligent Computing, pp. 101-108. Springer, Singapore, 2022.