Improving the Fisher Kernel for Large-Scale Image Classification

By Florent Perronnin, Jorge Sánchez, Thomas Mensink. 2010

2,882 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Very Deep Convolutional Networks for Large-Scale Image Recognition (2014, methodology): “In particular, an important role in t he advance of deep visual recognition architectures has been played by the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) [1], which has served as a testbed for a few generations of large-s cal image classification systems, from high-dimensional shal”
ImageNet Large Scale Visual Recognition Challenge (2014, methodology): “Im a g eN et L a rg e S ca le V isu a l R eco g n itio n C h a llen g e 1 7 ILSVRC 2010 Codename CLS LOC Insitutions Contributors and references Hminmax 54.4 Massachusetts Institute of Technology Jim Mutch, Sharat Chikkerur, Hristo Paskov, Ruslan Salakhutdinov, Stan Bileschi, Hueihan Jhuang IBM 70.1”
Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition (2014, methodology): “Then we still use win = da/ne and str = ba/nc to implement each pyramid pooling level.”
Do ImageNet Classifiers Generalize to ImageNet? (2019, background): “, 2011; Rahimi and Recht, 2009) and Fisher vectors for ImageNet (Perronnin et al., 2010).”
Describing Textures in the Wild (2013, methodology): “However, neither work tested the standard IFV formulation [27], which is well tuned for object recognition, developing instead variations specialized for texture analysis.”
Delving into Egocentric Actions (2015, positive): “We further take a signed square root of its scalar components and normalize the vector with a unit l2 norm [26].”
Flickr30k Entities: Collecting Region-to-Phrase Correspondences for Richer Image-to-Sentence Models (2015, methodology): “Key to their performance are state-of-the-art features for images (deep activation features from VGG net [32]) and text (Fisher vector pooling [29] on word2vec vectors [27] of individual words).”
Two-Stream Convolutional Networks for Action Recognition in Videos (2014, methodology): “Recent improvements of trajectory-based hand-crafted representations include compensation of global (camera) motion [10, 16, 26], and the use of the Fisher vector encoding [22] (in [26]) or its deeper variant [23] (in [21]).”
Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps (2013, methodology): “0% localisation error using a fully-supervised algorithm based on the part-based models [6] and Fisher vector feature encoding [11].”
Learning Deep Structure-Preserving Image-Text Embeddings (2015, methodology): “To represent sentences and phrases, we primarily use the Fisher vector (FV) representation [36] as suggested by Klein et al.”
Return of the Devil in the Details: Delving Deep into Convolutional Nets (2014, methodology): “The details of this construction can be found in [23]; here we follow the notation of [3].”
Hollywood in Homes: Crowdsourcing Data Collection for Activity Understanding (2016, methodology): “Finally, we encode the distribution of local descriptors over the video with Fisher vectors [34].”
On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation (2015, methodology): “We stick to the formulation of Fisher vectors as given in [11].”
Genetic CNN (2017, neutral): “, GPU), the Convolutional Neural Networks (CNNs) [19][32] have shown superior performance over the conventional Bag-ofVisual-Words [3][38][29] and compositional models [9].”
PKU-MMD: A Large Scale Benchmark for Continuous Multi-Modal Human Action Understanding (2017, methodology): “Densely tracking points in the optical flow field with more features like Histogram of Oriented Gradient (HOG), Histogram of Flow (HOF) and Motion Boundary Histograms (MBH) encoded by Fisher Vector [24, 42] achieved a good performance.”
Beyond Gaussian Pyramid: Multi-skip Feature Stacking for action recognition (2014, background): “[38, 40], together with the Fisher Vector encoding [26] yields the current state-of-theart performances on several benchmark action recognition datasets.”
Convolutional Two-Stream Network Fusion for Video Action Recognition (2016, positive): “Indeed the current state of the art performance [30, 34] on standard benchmarks such as UCF101 [24] and HMDB51 [13] is achieved by a combination of ConvNets and a Fisher Vector encoding [20] of hand-crafted features (such as HOF [14] over dense trajectories [33]).”