Classification of white flowers using transfer learning in convolutional neural networks

The role of flowers in pollination and balancing the plant world is vital. Their colors and fragrances attract pollinators such as bees, butterflies, and moths, with white flowers being especially attractive due to their fragrance, especially at night when they are perceived as purple by insects. The advancements in image processing using Convolutional Neural Networks (CNNs) have made it easier to detect and recognize plant diseases. The same technology can be used to classify different species of flowers based on color, which can improve pollination. In this study, images of eight types of white flowers, including Jasmine, Phlox, Leucanthemum Maximum, Cherry, Viola, Lily of the Valley, Apple tree, and Snowdrop, were used to train and test different convolutional network models. The study utilized transfer learning from pre-trained weights of various CNN models such as MobileNet, ResNet152V2, InceptionV3, Xception, and DenseNet201. The results showed that the MobileNet model had the highest accuracy at 95.83%, with a loading time of 1.39 seconds and a weight size of 16MB. Meanwhile, the DenseNet201 model had the lowest accuracy of 80.20%, with a loading time of 6.145 seconds and a weight size of 80MB.