Abstract:Significant progress has been made in instance image retrieval based on deep feature aggregation. However, existing approaches are limited by two issues: 1) The inability of deep features to localize target objects generates inaccurate feature descriptions and 2) using short vector feature representations provides unsatisfactory retrieval performance. To address these issues, we propose the compact deep semantic correlation descriptor (DSCD) approach, which has three main highlights: (1) Unlike manual labeling approaches, we propose a channel semantic correlation learning method to localize target objects. This method can learn semantic correlation information between channels from an auxiliary dataset and use this information as a priori knowledge to locate target objects well. (2) We propose a hierarchical attention mechanism to integrate multi-level features within target regions. It utilizes object and focus attention modules to capture global and local features, and constructs channel attention coefficients to enhance channels containing important target features. (3) We propose a general, yet efficient dimensionality reduction method named adaptive PCA-whitening to improve short vector feature retrieval performance. This method can dynamically fuse the feature vectors of different dimensions via adaptive coefficients, yielding a more compact and robust representation. Extensive experiments on six benchmark datasets show that our method achieves better retrieval performance compared to existing state-of-the-art unsupervised methods. Compared with the next-best unsupervised methods, the mAP scores of our method (dim = 128) are 11.2 %, 6.1 %, 9.3 %, 8.3 %, 14.3 %, and 22.1 % higher on the Oxford5K, Paris6K, Oxford105K, Paris106K, ROxford, and RParis datasets, respectively.

Dominant Sets Clustering for Image Retrieval.

Image retrieval based on incremental subspace learning

Image Retrieval Based on Dominant Set Clustering and Support Vector Machine

Incremental semi-supervised subspace learning for image retrieval.

Content-Based Image Retrieval By Dominant Color Feature

Merging Dominant Sets and DBSCAN for Robust Clustering and Image Segmentation

Algorithm about Image Retrieval Based on Dominant Color Set

DSET++: A robust clustering algorithm

Support Vector Machine Learning for Image Retrieval

An Efficient Hierarchical Clustering Algorithm and Its Application for Image Retrieval

Computer Image Content Retrieval Considering K-Means Clustering Algorithm

Discriminative Semantic Subspace Analysis for Relevance Feedback.

Discriminative Extreme Learning Machine To Content-Based Image Retrieval With Relevance Feedback

Image retrieval using compact deep semantic correlation descriptors

Robust Clustering Based on Dominant Sets

A Novel Approach Supporting Multi-image Queries in Image Retrieval

Efficient Image Retrieval Using Hierarchical K-Means Clustering

A Color Indexing Scheme Using Two-level Clustering Processing for Effective and Efficient Image Retrieval.

Using SVM to Organize the Image Database

A Effective Clustering Indexing Based On Emd For Web Image Retrieval

Re-ranking Algorithm Using Clustering and Relevance Feedback for Image Retrieval