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.

Combining Self-Aggregation Vector and Indicator Matrix for Improved Object Retrieval

Aggregating Hierarchical Binary Activations for Image Retrieval

Exploring Geometric Information in CNN for Image Retrieval.

Learning Feature Embedding with Strong Neural Activations for Fine-Grained Retrieval

Deep Image Retrieval: Indicator and Gram Matrix Weighting for Aggregated Convolutional Features

Adaptive Co-weighting Deep Convolutional Features For Object Retrieval

Unsupervised Semantic-Based Convolutional Features Aggregation for Image Retrieval

Image retrieval based on aggregated deep features weighted by regional significance and channel sensitivity

Image retrieval using compact deep semantic correlation descriptors

Co-weighting semantic convolutional features for object retrieval.

Exploring Spatial and Channel Contribution for Object Based Image Retrieval

Fast Object Retrieval Using Direct Spatial Matching

Learning Token-Based Representation for Image Retrieval

End-to-end Semantic-Aware Object Retrieval Based on Region-Wise Attention

Deepindex For Accurate And Efficient Image Retrieval

Scalable Object Retrieval with Compact Image Representation from Generic Object Regions

Attention-based Global Feature Extraction Method for Image Retrieval

Deep Aggregation of Regional Convolutional Activations for Content Based Image Retrieval

Combining Convolutional Neural Network and Markov Random Field for Semantic Image Retrieval

Deep-seated features histogram: A novel image retrieval method

Building Discriminative CNN Image Representations for Object Retrieval Using the Replicator Equation.