Improving Bag-of-Features for Large Scale Image Search

By Herv'e J'egou, Matthijs Douze, Cordelia Schmid. 2010

789 citations recorded by the source corpus.

Subfields: Aerospace Engineering, 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 Aerospace Engineering

In Computer Vision and Pattern Recognition

In Information Retrieval & Search

Citations in the corpus

Aggregating Local Image Descriptors into Compact Codes (2012, methodology): “For K 1⁄4 4;096 we obtain a mAP 1⁄4 68:9%, which outperforms any result reported for standard BOW on this data set ([35] reports mAP 1⁄4 57:2% with a 200k vocabulary using the same descriptors).”
Particular object retrieval with integral max-pooling of CNN activations (2015, methodology): “We additionally use 100k Flickr images (Philbin et al., 2007) to compose Oxford105k and Paris106k, respectively.”
End-to-End Learning of Deep Visual Representations for Image Retrieval (2016, methodology): “Encoding techniques, such as the Fisher Vector (Perronnin and Dance, 2007; Perronnin et al, 2010) or the VLAD descriptor (Jégou et al, 2010) have been used for example by Perronnin et al (2010); Gordo et al (2012); Jégou and Chum (2012); Radenovic et al (2015).”
Working hard to know your neighbor's margins: Local descriptor learning loss (2017, methodology): “In our case (HardNet+–HQE), a visual vocabulary of 65k visual words is learned, with additional Hamming embedding (HE) [44] technique that further refines descriptor assignments with a 128 bits binary signature.”
Hough Transform and 3D SURF for Robust Three Dimensional Classification (2010, methodology): “Following standard practice [19, 20] in large-scale image searching, we set the number of visual words (clusters) to 10% of the total number of features in our training set.”
Time-series classification using neural Bag-of-Features (2017, methodology): “, [10]) used clustering algorithms, such as k-means, to learn generic dictionaries that minimize the reconstruction loss of the quantized features.”
Sliced Wasserstein Kernels for Probability Distributions (2015, methodology): “This is for instance the case for histogram-based methods like the Bag-of-Words (BoW) [21], feature matching [17], co-occurence matrices in texture analysis [15], action recognition [43], and many more.”
Classifying Compound Structures in Satellite Images: A Compressed Representation for Fast Queries (2015, methodology): “The assumption of using selected visual words for performing a query/classiﬁcation suggests the use of an inverted ﬁle [3] for storing in a compact form the compound structure representation.”
Good Practice in CNN Feature Transfer (2016, neutral): “Holidays [12], Ukbench [25], and Oxford5k [29].”
Asymmetric metric learning for knowledge transfer (2020, background): “se, or not at all [12, 17, 34, 32, 45, 10]. In image retrieval, there are efforts to reduce the asymmetry of k-nearest neighbor relations [35], or use asymmetry to mitigate the effect of quantization [33], or handle partial similarity [84] or alignment [66]. In classiﬁcation, it is common to use asy”
Dynamic Match Kernel With Deep Convolutional Features for Image Retrieval (2018, neutral): “[51] [52] [18] [56] [53] [57] [58] [59] [60] [61] [62] [63]”
Large Scale Nearest Neighbors Search Based on Neighborhood Graph (2013, background): “Finding the k-nearest neighbors (KNN) of a single query point in an extremely large collection of high-dimensional vectors is a fundamental problem in multimedia retrieval such as image retrieval [1, 2, 3], computer vision [4], music retrieval [5], etc.”