TEMU KEMBALI CITRA MENGGUNAKAN FITUR PRE-TRAINED CONVOLUTIONAL NEURAL NETWORK

As the world of digital technology continues to develop, there is more and more data, especially image data. One branch of computer vision that extracts features from images, namely Content Based Image Retrieval, is a solution for image retrieval and is useful for various fields. Using the pre-trained VGG16 method to produce precision and recall values. Using the Corel-1k dataset of 1000 image data which is processed in the initial stage, namely preprocessing, feature extraction using a pre-trained model, measuring the similarity distance between the database image and the query image and evaluation. In previous research, using the same dataset with different methods produced an average precision value of 74% and recall of 64%. Then the results obtained from this research using the same dataset and using the VGG16 pre-trained method produced an average precision value of 83%. and an average recall of 82%.