KLASIFIKASI CITRA BUNGA MENGGUNAKAN MODEL PRE-TRAINED MOBILENET

Flowers are part of plants that have very diverse and beautiful colors and shapes. A good understanding of flowers is essential to help in identifying new or rare species when they are discovered. Flowers classification is a difficult problem due to the high diversity of shapes, color distribution, lighting conditions, and exposure deformation. Another difficulty is finding color descriptors, shapes, patterns, and also suitable classifiers to select significant features. To Solving problems in classifying flower images can be done by various methods. One of them uses the Convolutional Neural Network method. This research aims to classify flower images using the Convolutional Neural Network method with the Mobilenet model to get better performance. In addition, this study aims to determine the effect of layer dropouts in overcoming overfitting problems. The evaluation results show that Mobilenet models without using dropout layers have the best accuracy results with values of 99.90% in data train and 90.93% in data validation. Apart from the accuracy results, it can be seen from the total time required for the Mobilenet model to carry out the model training process faster than the InceptionResnetV2 model. This research proves that layer dropouts cannot always solve the problem of overfitting the built model.