Supervised Hashing for Image Retrieval via Image Representation Learning

By Rongkai Xia, Yan Pan, Hanjiang Lai, Cong Liu, Shuicheng Yan. AAAI Conference on Artificial Intelligence, 2014

977 citations recorded by the source corpus.

Hashing is a popular approximate nearest neighbor search approach for large-scale image retrieval. Supervised hashing, which incorporates similarity/dissimilarity information on entity pairs to improve the quality of hashing function learning, has recently received increasing attention. However, in the existing supervised hashing methods for images, an input image is usually encoded by a vector of hand-crafted visual features. Such hand-crafted feature vectors do not necessarily preserve the accurate semantic similarities of images pairs, which may often degrade the performance of hashing function learning. In this paper, we propose a supervised hashing method for image retrieval, in which we automatically learn a good image representation tailored to hashing as well as a set of hash functions. The proposed method has two stages. In the first stage, given the pairwise similarity matrix $S$ over training images, we propose a scalable coordinate descent method to decompose $S$ into a product of $HH^T$ where $H$ is a matrix with each of its rows being the approximate hash code associated to a training image. In the second stage, we propose to simultaneously learn a good feature representation for the input images as well as a set of hash functions, via a deep convolutional network tailored to the learned hash codes in $H$ and optionally the discrete class labels of the images. Extensive empirical evaluations on three benchmark datasets with different kinds of images show that the proposed method has superior performance gains over several state-of-the-art supervised and unsupervised hashing methods.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/f6338f23860f9c4808586bbc7e8907d33836147f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Bit-Scalable Deep Hashing With Regularized Similarity Learning for Image Retrieval and Person Re-Identification (2015, neutral): “[35], which utilizes CNN for supervised hashing learning.”
Two-Stream Deep Hashing With Class-Specific Centers for Supervised Image Search (2020, background): “Convolutional neural networks-based hashing (CNNH) [52] decomposes the hash function learning into two stages.”
A Survey on Learning to Hash (2016, methodology): “The similarity-similarity difference minimization criterion is adopted in [158].”
Deep Supervised Hashing for Fast Image Retrieval (2016, methodology): “CNNH [29] 0.”
Feature Learning Based Deep Supervised Hashing with Pairwise Labels (2015, methodology): “Both DPSH and CNNH are deep hashing methods with pairwise labels.”
Deep Hashing Network for Efficient Similarity Retrieval (2016, methodology): “DHN-E is the DHN variant using widely-adopted pairwise squared loss L = sij ∈S s ij − 1 K z l i , z l j 2 (Liu et al. 2012; Xia et al. 2014) instead of the pairwise cross-entropy loss (5).”
Unsupervised Deep Hashing with Similarity-Adaptive and Discrete Optimization (2018, methodology): “CNNH [51] is a two-step hashing method, which learns a deep CNN based hash function with the pre-computed binary codes.”
Deep Visual-Semantic Hashing for Cross-Modal Retrieval (2016, methodology): “Meanwhile, latest deep hashing methods [40, 23, 46, 5] yielded state of art results on many datasets, but these methods are limited to single-modal retrieval.”
A Decade Survey of Content Based Image Retrieval Using Deep Learning (2020, methodology): “image retrieval [86], [89], [93], [135], semi-supervised image retrieval [115], cross-modal retrieval [143], [144], sketch based image retrieval [145], [146], object retrieval [147], [148], etc.”
High-Dimensional Sparse Cross-Modal Hashing with Fine-Grained Similarity Embedding (2021, methodology): “Particularly, the features of image modality used by HSCHcnn are replaced with deep features extracted from CNN-F model [53]”
Deep learning of binary hash codes for fast image retrieval (2015, methodology): “Moreover, it attains a precision of 89% while varying the number of retrieved images, improving the performance by a margin of 30% compared to CNNH+ [30].”
Comparative Deep Learning of Hybrid Representations for Image Recommendations (2016, methodology): “In view of the limitation of hand-crafted image features such as those designed in [4, 15, 20, 28], more and more research focuses on designing effective deep learning models to extract image representations automatically [17, 21, 31].”
Central Similarity Quantization for Efficient Image and Video Retrieval (2019, methodology): “The “deep learning to hash” methods such as CNNH [44], DNNH [14], DHN [48], DCH [1] and Hash-Net [2] have been successfully applied to image hashing.”
How should we evaluate supervised hashing? (2016, neutral): “Although the work of [22] was both deep and a SSH setup, their evaluation metric differs from our definition of mAP and thus we have not included them in our comparison.”
Deep semantic ranking based hashing for multi-label image retrieval (2015, methodology): “[33] learn hash functions using CNNs to fit the learned hash codes.”