Fisher Kernels on Visual Vocabularies for Image Categorization

By Florent Perronnin, Christopher R. Dance. 2007

1,802 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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

Deep Residual Learning for Image Recognition (2015, methodology): “In image recognition, VLAD [18] is a representation that encodes by the residual vectors with respect to a dictionary, and Fisher Vector [30] can be formulated as a probabilistic version [18] of VLAD.”
Going deeper with convolutions (2014, methodology): “Figure 3: Precision and success plots of overall performance comparison for the 51 videos in the benchmark [8] (best-viewed on high-resolution display).”
ImageNet Large Scale Visual Recognition Challenge (2014, methodology): “The honorable mention XRCE team (Perronnin et al., 2010) used an improved Fisher vector representation (Perronnin and Dance, 2007) along with PCA dimensionality reduction and data compression followed by a linear SVM. Fisher vector-based methods have evolved over ﬁve years of the challenge and…”
Aggregating Local Image Descriptors into Compact Codes (2012, methodology): “The last two are related [11], as search efﬁciency can be approximated by the amount of memory visited during the search.”
Group Normalization (2018, methodology): “2010) and Fisher Vectors (FV) (Perronnin and Dance 2007) are also group-wise features where a group can be thought of as the sub-vector computed with respect to a cluster.”
Understanding deep image representations by inverting them (2014, background): “Most image understanding and computer vision methods build on image representations such as textons [15], histogram of oriented gradients (SIFT [18] and HOG [4]), bag of visual words [3][25], sparse [35] and local coding [32], super vector coding [37], VLAD [9], Fisher Vectors [21], and, lately, dee”
Deep Multimodal Feature Analysis for Action Recognition in RGB+D Videos (2016, background): “BOvW [58] and FV [43] or Fourier temporal pyramids in [73].”
Bilinear CNN Models for Fine-Grained Visual Recognition (2015, methodology): “linear ﬁlter-bank responses (e.g., wavelets and steerable pyramids). The use of second-order features of ﬁlter-bank responses was pioneered by Portilla and Simoncelli [50]. Recent variants such as FV [46] and O2P [8] with SIFT were shown to be a highly effective for image classiﬁ- cation and semanti”
Visualizing Deep Convolutional Neural Networks Using Natural Pre-images (2015, background): “Notable examples of representations include textons [25], histogram of oriented gradients (SIFT [28] and HOG [5]), bag of visual words [4][39], sparse [49] and local coding [46], super vector coding [53], VLAD [17], Fisher Vectors [34], and, lately, ∗aravindh@robots.ox.ac.uk †vedaldi@robots.ox.ac.uk”
Fisher Vectors Derived from Hybrid Gaussian-Laplacian Mixture Models for Image Annotation (2014, methodology): “Then the Fisher Vector of the LMM is (see supplementary material): As in [32], the diagonal of the Fisher Information Matrix F is approximated in order to normalize the dynamic range of the different dimensions of the gradient vectors.”
Multi-scale Orderless Pooling of Deep Convolutional Activation Features (2014, methodology): “For this, we adopt Vectors of Locally Aggregated Descriptors (VLAD) [16,25], which are a simplified version of Fisher Vectors (FV) [15].”
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).”
Learnable pooling with Context Gating for video classification (2017, methodology): “Finally, for Fisher Vector encoding, we modify the NetVLAD architecture to allow learning of second order feature statistics within the clusters.”