Vegetable Image Retrieval with Fine-tuning VGG Model and Image Hash

Image descriptors based on activations of Convolutional Neural Networks (CNN) have become dominant in image retrieval due to their discriminative power, compactness of the representation, and the efficiency of search. Fine-tune existing CNN models for image retrieval in specific domain is significant for content-based image retrieval tasks. Inspired by recent successes of CNN with hierarchical features, in this paper, we fine-tuning VGG model to learn features for special vegetable dataset with the classification task. Furthermore, we propose utilizing some PCA Hashing strategies combinate CNN features extracted by the fine-tuned model to improve the performance of special domain CBIR tasks. Our experimental results demonstrate that leveraging the method we proposed can improve the performance of CBIR and the mAP increased by 10 to 20 percent in seam Hash code bits, compared to the model before fine-tuning. (C) 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.