Total recall II: Query expansion revisited

By Ondřej Chum, Andrej Mikulík, Michal Perdoch, Jiri Matas. Computer Vision and Pattern Recognition, 2011

318 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/b368f2df712c94f77b913a67927cb88b84b10ba2

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

Visual Instance Retrieval with Deep Convolutional Networks (2014, methodology): “There are also further tricks and techniques that can be exploited from the established literature in image retrieval such as query expansion (Chum et al., 2011) and spatial re-ranking (Sivic & Zisserman, 2003).”
Deep Image Retrieval: Learning Global Representations for Image Search (2016, methodology): “All these methods can be combined with other post-processing techniques such as query expansion [29,30,31].”
CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples (2016, background): “Retrieval systems have reached a higher level of maturity by incorporating large visual codebooks [3,4], spatial verification [3,5] and query expansion [6,7,8].”
Ranking consistency for image matching and object retrieval (2014, background): “The results of incremental spatial verification (iSP) are cited from Total Recall II [44].”
Particular object retrieval with integral max-pooling of CNN activations (2015, background): “Query expansion approaches signiﬁcantly increase the performance (Chum et al., 2011), at the cost of larger query times.”
Three things everyone should know to improve object retrieval (2012, methodology): “The query expansion methods of [5, 6], where BoW vectors from spatially verified regions are used to issue new queries, address the problem of feature detection drop out in addition to quantization and noise on the descriptor, and again improves recall.”
Patch-NetVLAD: Multi-Scale Fusion of Locally-Global Descriptors for Place Recognition (2021, background): “Place matches obtained through global descriptor matching are often re-ranked using sequential information [26, 85, 49], query expansion [31, 13], geometric verification [41, 27, 52] and feature fusion [83, 86].”
Fine-Tuning CNN Image Retrieval with No Human Annotation (2017, background): “Impressively, in this work we show that CNNs dominate the image search task by outperforming state-of-the-art methods that have reached a higher level of maturity by incorporating large visual codebooks [33], [34], spatial verification [33], [35] and query expansion [36], [37], [38].”
MatchMiner: Efficient Spanning Structure Mining in Large Image Collections (2012, methodology): “Incremental spatial re-ranking [11] automatically detects tf-idf failure (too few features landing in the same visual words are geometrically consistent), incrementally builds a statistical model of the query object, and learns relevant spatial context to boost retrieval performance.”
Cross-Dimensional Weighting for Aggregated Deep Convolutional Features (2015, background): “When further combining our approach with query expansion, we get even better results, that compare to (or surpass in Paris) far more sophisticated approaches like [21, 5, 33] that are based on local features and include spatial verification steps.”
The knowing camera 2: recognizing and annotating places-of-interest in smartphone photos (2014, background): “The approaches vary widely from query expansion[7, 6] to soft assignment[17, 10], all aiming at increasing recall while maintaining high precision.”
Revisiting Oxford and Paris: Large-Scale Image Retrieval Benchmarking (2018, methodology): “Numerous methods of image retrieval [7, 31, 5, 27, 47, 3, 48, 20, 37, 10] and visual localization [9, 1] have used these datasets for evaluation.”
SIFT Meets CNN: A Decade Survey of Instance Retrieval (2016, methodology): “For other QE variants [101], [33], the proposed improvements only add marginal cost compared to performing another search, so their complexity is similar to basic QE methods.”
InLoc: Indoor Visual Localization with Dense Matching and View Synthesis (2019, background): “[25], [37], [39], [58], [75], [86] and can be further improved by spatial re-ranking [60], informative feature selection [24], [25] or feature weighting [31], [38], [65], [82].”
All About VLAD (2013, methodology): “clustering on SIFTs): (i) Paris 6k [14], which is analogous to the Oxford buildings dataset, and is often used as an independent dataset from the Oxford buildings [1, 3, 5, 14]; (ii) Flickr60k [6], which contains 60k images downloaded from Flickr, and is used as an independent dataset from the Holid”
Google Landmarks Dataset v2 – A Large-Scale Benchmark for Instance-Level Recognition and Retrieval (2020, methodology): “Other re-ranking techniques which are especially important for retrieval tasks, such as query expansion (QE) [17, 44] and graph traversal [13], were also employed.”
Dynamic Match Kernel With Deep Convolutional Features for Image Retrieval (2018, unknown): “[64] [51] [1] [26] [18] [53] [65] [66] [12] [52] [67] [58] [62] [19]”
Visual Place Recognition with Repetitive Structures (2015, background): “Examples include: (i) learning better visual vocabularies [21, 28]; (ii) developing quantization methods less prone to quantization errors [14, 27, 44]; (iii) combining returns from multiple query images depicting the same scene [4, 6]; (iv) exploiting the 3D or graph structure of the database [11, ”