Total Recall: Automatic Query Expansion with a Generative Feature Model for Object Retrieval

By Ondřej Chum, James Philbin, Josef Sivic, Michael Isard, Andrew Zisserman. 2007

986 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Information Retrieval & Search

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 Information Retrieval & Search

Citations in the corpus

Unsupervised Visual Representation Learning by Context Prediction (2015, methodology): “Because our features are more semantically-tuned, we can use a much weaker type of geometric verification than [7].”
IM2GPS: estimating geographic information from a single image (2008, background): “This is unlikely, not just because of computational cost, but simply because the set of all existing photographs is still not large enough to exhaustively sample the entire world.”
MultiGrain: a unified image embedding for classes and instances (2019, neutral): “Possible improvement include reﬁnement steps such as geometric veriﬁcation [33], query expansion [4, 42], or database-side pre-processing or augmentation [41, 44].”
XCiT: Cross-Covariance Image Transformers (2021, methodology): “All features have a constant stride of 8 or 16 based on the patch size, and the feature resolutions are adjusted to have strides of [4, 8, 16, 32], similar to ResNet-FPN backbones, where the downsampling is achieved by max pooling and the upsampling is obtained using a single transposed convolution ”
End-to-End Learning of Deep Visual Representations for Image Retrieval (2016, methodology): “All these methods can be combined with postprocessing techniques such as query expansion (Chum et al, 2007, 2011; Arandjelovic and Zisserman, 2012).”
Near Duplicate Image Detection: min-Hash and tf-idf Weighting (2008, background): “Recently, most of the successful image indexing approaches are based on the bag-ofvisual-words representation [5, 9, 17, 18, 19].”
Building Rome on a Cloudless Day (2010, methodology): “[2] parallelize the matching process and use approximate nearest neighbor search and query expansion [21] on a cluster of 62 machines each one comparable to our single PC.”
Evaluation of GIST descriptors for web-scale image search (2009, background): “Re-ranking based on the estimation of a geometrical transform [17], as well as query expansion [1] –inspired by text retrieval–, improves the accuracy.”
Appearance-only SLAM at large scale with FAB-MAP 2.0 (2011, background): “Also of direct relevance is the research on content-based image retrieval systems in the computer vision community, where systems have been described that deal with more than a million images (Philbin et al. 2007; Chum et al. 2007; Nistér and Stewenius 2006; Jégou et al. 2008).”
Reconstructing the World* in Six Days *(As Captured by the Yahoo 100 Million Image Dataset) (2015, methodology): “To improve basic retrieval performance, they proposed to leverage query expansion [7] to obtain a greater number of overlapping images.”
Unifying Deep Local and Global Features for Image Search (2020, methodology): “ing ties using the global feature distance. Our focus is on improving global and local features for image retrieval, so we do not consider techniques that post-process results such as query expansion [10,45] or diffusion/graph traversal [22,8]. These are expensive due to requiring additional passes ”
Learning and Calibrating Per-Location Classifiers for Visual Place Recognition (2013, background): “Examples include: (i) learning better visual vocabularies from training examples with matched/non-matched descriptors [19, 23]; (ii) developing quantization methods less prone to quantization er-1 rors [11, 22] or (iii) combining results from multiple query images depicting the same scene [4, 5].”
Thinking Fast and Slow: Efficient Text-to-Visual Retrieval with Transformers (2021, background): “Query expansion [9] is another re-ranking technique where the query is reformulated given top retrieved candidates, and recent work has brought attention mechanisms into deep learning methods for query expansion [21].”
Negative evidences and co-occurrences in image retrieval: the benet of PCA and whitening (2012, background): “A majority of the papers rely on the bag-of-words (BOW) representation [1, 5, 2, 3] or its derivatives, e.”
Visual landmark recognition from Internet photo collections: A large-scale evaluation (2014, methodology): “se problems has already been studied in isolation, so far there has not been a systematic evaluation of all three aspects in the context of a fully automatic pipeline. Image retrieval approaches like [7,8] ﬁnd matching images for a query, but have no notion of the semantics of the depicted content. ”
Visual Instance Retrieval with Deep Convolutional Networks (2014, methodology): “For a fair comparison, the benchmarking of our methods will be with pipelines that do not include any search reﬁnement procedures such as query expansion 17) and spatial re-ranking 18)19) .”
Ranking consistency for image matching and object retrieval (2014, background): “For example, query expansion [7] adds more relevant features from an automatic scheme of sampling selection; Bayesian relevance feedback [15] needs users to identify retrieved images as being relevant or not, and then adjusts the query by Bayesian decision theory; Trademark retrieval [16]…”