Large-scale image retrieval with compressed Fisher vectors

By Florent Perronnin, Yan Liu, Jorge Sánchez, Hervé Poirier. 2010

864 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): “A low dimensional vector can be indexed efﬁciently, but its discriminative power may be too low.”
Deep Structured Output Learning for Unconstrained Text Recognition (2014, methodology): “Rodriguez-Serrano et al. (2013) use aggregated Fisher Vectors (Perronnin et al. (2010)) and a Structured SVM framework to create a joint word-image and text embedding.”
CNN Features Off-the-Shelf: An Astounding Baseline for Recognition (2014, methodology): “In this section we compare the CNN representation to the current state-of-the-art representations of retrieval including VLAD[4], BoW 200k, IFV[31] and BoB[3].”
Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition (2014, methodology): “[24] use aggregated Fisher Vectors [23] and a Structured SVM framework to create a joint word-image and text embedding.”
Learning Fine-Grained Image Similarity with Deep Ranking (2014, methodology): “5% 2863 Fisher [20] 67.”
Fisher Vectors Derived from Hybrid Gaussian-Laplacian Mixture Models for Image Annotation (2014, methodology): “Some of the most widely used improvements were introduced by Perronnin et al. [33].”
NetVLAD: CNN Architecture for Weakly Supervised Place Recognition (2015, methodology): “While there have been many improvements in designing better image retrieval [3, 4, 12, 13, 18, 25, 26, 27, 29, 32, 48, 51, 52, 53, 54, 70, 77, 78, 81] and place recognition [5, 10, 11, 16, 17, 24, 35, 44, 46, 62, 63, 64, 74, 79, 80] systems, not many works have performed learning for these tasks.”
Reading Text in the Wild with Convolutional Neural Networks (2014, methodology): “[51] use aggregated Fisher Vectors [47] and a Structured SVM framework to create a joint word-image and text embedding.”
Multi-scale Orderless Pooling of Deep Convolutional Activation Features (2014, methodology): “Note that applying PCA after the two stages (local patch activation and global pooled vector) is a standard practice in previous works [26,27].”
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).”
Detection and Retrieval of Out-of-Distribution Objects in Semantic Segmentation (2020, background): “(5)) which can be beneficial for high dimensional data [26] like the feature vectors extracted from neural networks,”
CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples (2016, methodology): “Representative approaches are Fisher vectors [9], VLAD [10] and alternatives [11,12,13].”
Where to Buy It: Matching Street Clothing Photos in Online Shops (2015, methodology): “Many recent retrieval methods at a high-level consist of three main steps: pooling local image descriptors (such as Fisher Vectors [25] [27] [26] or VLAD [15]), dimensionality reduction, and indexing.”
All About VLAD (2013, methodology): “For the Holidays dataset we outperform the best method (improved Fisher vectors [12]) by 3.”
Label embedding for text recognition (2013, background): “This is an approximate explicit embedding of the Fisher kernel, which obtained state-of-the-art results in image retrieval [20].”
From single image query to detailed 3D reconstruction (2015, methodology): “Most of the approaches pose the problem as a nearest neighbor search in a descriptor space, such as bag-of-words [28, 21, 25, 11], VLAD [12, 2], Fischer vectors [24], or exhaustive matching [34, 27].”
DeepHash: Getting Regularization, Depth and Fine-Tuning Right (2015, methodology): “However, the dimensionality of such descriptors is typically very high: 8192 to 65536 ﬂoating point numbers for FVs[1] and 4096 for DCNNs [2].”