A New Paphiopedilum Orchid Database and Its Recognition Using Convolutional Neural Network

This paper discusses a visual recognition system, for identifying the Pa-phiopedilum orchid, often called the Venus slipper. The dataset consists of 100 sample images for each of 15 species of orchid, for a total of 1500 images. All the images of this dataset were taken at the Paphiopedilum orchid gardens and manually classified by experts. This work also implemented a recognition system based on a deep learning approach using a combination of convolutional neural network (CNN) and the Inception-v3 feature extractor of the TensorFlow platform. The implemented recognition system can deliver recognition rates of up to 98.6%, demonstrating excellent recognition performance by the CNN model. Finally, we demonstrate a prototype orchid recognition system, implemented as an Android mobile application.