Abstract
Deep hashing, the combination of advanced convolutional neural networks and efficient hashing, has recently achieved impressive performance for image retrieval. However, state-of-the-art deep hashing methods mainly focus on constructing hash function, loss function and training strategies to preserve semantic similarity. For the fundamental image characteristics, they depend heavily on the first-order convolutional feature statistics, failing to take their global structure into consideration. To address this problem, we present a deep covariance estimation hashing (DCEH) method with robust covariance form to improve hash code quality. The core of DCEH involves covariance pooling as deep hashing representation, performing global pairwise feature interactions. The covariance pooling can capture richer statistic information of deep convolutional features and produce more informative global representations.Due to convolutional features are usually high dimension and small sample size, we estimate robust covariance by shrinking its eigenvalues using power normalization, forming an independent structural layer. Then the structural layer is embedded into deep hashing paradigm in an end-to-end learning manner. Extensive experiments on three benchmarks show that the proposed DCEH outperforms its counterparts and achieves superior performance.
Highlights
With the explosive growth of image and video data on the web, near neighbor (NN) search has been widely used in information retrieval, computer vision and other fields
(3) Experiments on several benchmarks have demonstrated the effectiveness of deep covariance estimation hashing (DCEH) and achieved competitive performance compared with other deep hashing methods only using the first-order conv. feature statistics
EXPERIMENTS we carry out extensive experiments to evaluate the effectiveness of the proposed DCEH and compare it with other approaches on three benchmarks, in which the images are in a wide spectrum of image types including handwritten digits of MNIST [41], tiny objects of CIFAR-10 [42], as well as web images of NUS-WIDE [43]
Summary
With the explosive growth of image and video data on the web, near neighbor (NN) search has been widely used in information retrieval, computer vision and other fields. Exemplar supervised hashing methods include deep supervised hashing with triplet labels (DTSH) [18], network in network hashing (NINH) [19], deep pairwise supervised hashing (DPSH) [20], asymmetric deep supervised hashing (ADSH) [33], deep hashing network (DHN) [22] and recent generative adversarial networks (GANs) based deep hashing [27], [28] These methods employ deep convolutional (conv.) features with extra meaningful techniques, they all ignore modeling the global pairwise feature interactions (the second-order statistics), which could lead to more promising performance. We propose a novel deep covariance estimation hashing (DCEH) method, which can generate compact hash code by establishing global statistics of feature interactions in an end-to-end learning framework. On covariance estimation can be learned under the pairwise similarity constraint It collects global statistics, by which the correlation between features can be explored to generate more powerful hash representation. By which the correlation between features can be explored to generate more powerful hash representation. (3) Experiments on several benchmarks have demonstrated the effectiveness of DCEH and achieved competitive performance compared with other deep hashing methods only using the first-order conv. feature statistics
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
