DEEP CONVOLUTIONAL NEURAL NETWORKS FEATURES FOR IMAGE RETRIEVAL

Content-Based Image Retrieval (CBIR) has become one of the trending areas of research in computer vision. In traditional CBIR, the features are considered as hand-crafted features. The state-of-the-art technology for feature extraction is to use deep convolutional neural networks (CNN). In this paper, four deep convolutional neural networks (CNNs): AlexNet, VGG-16, GoogleNet, and ResNet-101 with transfer learning are used to extract and the features from the image. By using these features, dl-distance is used to compare the query images with the images in the image dataset. To evaluate the efficiency of these four models, five standard performance measures are calculated i.e., Average Precision Rate (APR), Average Recall Rate (ARR), F-Measure, Average Normalized Modified Retrieval Rank (ANMRR) and Total Minimum Retrieval Epoch (TMRE). Six benchmark image datasets: Corel-1K, Corel-5K, Corel-10K, VisTex, STex, and Color Brodatz are used to corroborate the performance of the four CNN models for CBIR.