Iterative Quantization: A Procrustean Approach to Learning Binary Codes for Large-Scale Image Retrieval

By Yunchao Gong, Svetlana Lazebnik, Albert Gordo, Florent Perronnin. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013

1,709 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/f447d9de5b62d40345a5da90c60bbc47c967baab

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

In Media Technology

Citations in the corpus

Learning Compact Binary Descriptors with Unsupervised Deep Neural Networks (2016, neutral): “Linear Discriminat Analysis (LDA) is also applied to learn binary descriptors [14] [38].”
Learning With Average Precision: Training Image Retrieval With a Listwise Loss (2019, methodology): “, SIFT [36]), aggregated using bag-ofwords representations [14] or more elaborate schemes [16, 19, 55, 30], to obtain image-level signatures that could then be compared to one another in order to find their closest matches.”
Supervised Discrete Hashing (2015, methodology): “The proposed method is compared against several state-of-the-art supervised hashing methods including BRE [13], MLH [24], SSH [34], CCA-ITQ [10], KSH [20], FastHash [17] and unsupervised methods including PCA-ITQ [10], AGH [21] and IMH [29] with t-SNE [33].”
Hashing with binary autoencoders (2015, methodology): “Using ITQ (Gong et al., 2013) produces also good results (occasionally better but generally somewhat worse than AGH), and is a simpler and faster option if so desired.”
Hashing for Distributed Data (2015, methodology): “In other words, if the dictionary C can also be updated in the distributed setting, then the whole learning process of hashing can be achieved in a completely distributed manner.”
Fracking Deep Convolutional Image Descriptors (2014, methodology): “This line of works includes unsupervised techniques based on hashing as well as supervised approaches using Linear Discriminant Analysis (Brown et al., 2011; Gong et al., 2012; Strecha et al., 2012), boosting (Trzcinski et al., 2013), and convex optimization (Simonyan et al., 2014).”
Binary Multi-View Clustering (2019, background): “cal requirement of binary code learning [32].”
Image Matching from Handcrafted to Deep Features: A Survey (2020, methodology): “In this case, a plethora of methods have been proposed (Kulis and Darrell 2009; Wang et al. 2010; Strecha et al. 2012; Liu et al. 2012a; Norouzi and Blei 2011; Gong et al. 2013; Shakhnarovich 2005), among which image matching is considered an important experimental validation task.”
Compressing Deep Convolutional Networks using Vector Quantization (2014, methodology): “Many learning-based binarization or product quantization methods are available, such as Spectral Hashing (Weiss et al., 2008), Iterative Quantization (Gong et al., 2012), and Catesian k means (Norouzi & Fleet, 2013), among others.”
Discrete Collaborative Filtering (2016, methodology): “Similarly, Zhou and Zha [33] rotated their learned features by running ITQ [9] to generate hash codes.”
A Survey on Learning to Hash (2016, methodology): “Iterative quantization [35], [36] preprocesses the centralized data, by reducing the dimension using PCA toM dimensions, v 1⁄4 P>x, where P is amatrix of size d M (M4 d) computed using PCA, and then finds an optimal rotation R followed by a scalar quantization.”
Central Similarity Quantization for Efficient Image and Video Retrieval (2019, methodology): “Baselines and evaluation metrics We compare retrieval performance of our proposed CSQ with nine classical or state-of-the-art hashing/quantization methods, including four supervised shallow methods ITQ-CCA [7], BRE [13], KSH [24], SDH [32] and five supervised deep methods CNNH [44], DNNH [14], DHN [”
Efficient Nearest Neighbor Language Models (2021, methodology): “The tool we use for PCA, the FAISS PCA implementation, applies random rotation to the PCA output vectors by default to re-balance variances of components of a vector (Gong et al., 2012), which may provide additional benefits over vanilla PCA on product vector quantization inside the index.”
Accelerating Large-Scale Inference with Anisotropic Vector Quantization (2019, methodology): “ted codebook learning strategies, such as [17–19] for binary quantization and [20, 8] for product quantization. There are also lines of work that focus on learning transformations before quantization [21, 22]. Different from these methods which essentially minimize reconstruction error of the databa”
Unsupervised Semantic-Preserving Adversarial Hashing for Image Search (2019, methodology): “The proposed method is compared with ﬁve state-of-the-art unsupervised shallow architecture-based hashing methods (ITQ, SH, DSH, SpH, and SGH) and two recently proposed deep unsupervised hashing methods (DeepBit and BGAN).”
Learning in high-dimensional multimedia data: the state of the art (2015, background): “64 bits 128 bits BRE [44] High low SSH [92] low low ITQ [23] fair fair KSH [53] High fair FastHash [50] High There are also some further developments extended from the standard hashing algorithms: Multiple Feature Hashing (MFH) [81] learns a group of hash functions to generate binary codes for multi”
Discriminative Learning of Deep Convolutional Feature Point Descriptors (2015, methodology): “This line of works includes unsupervised techniques based on hashing as well as supervised approaches using Linear Discriminant Analysis [3, 9, 24], boosting [29], and convex optimization [23].”
Jointly Optimizing Query Encoder and Product Quantization to Improve Retrieval Performance (2021, methodology): “Unsupervised compression baselines include LSH [24], ITQ+LSH [19], PQ [25], OPQ [18], ScaNN [20], OPQ+ScaNN, and HNSW+OPQ [26].”