Large-scale image categorization with explicit data embedding

By Florent Perronnin, Jorge Sánchez, Yan Liu. 2010

182 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

Aggregating Local Image Descriptors into Compact Codes (2012, methodology): “for GMM modeling and in [31], [32] for better separability”
A Multi-View Embedding Space for Modeling Internet Images, Tags, and Their Semantics (2012, methodology): “All the other descriptors above are histograms, and for them we adopt the exact Bhattacharyya kernel mapping given by term-wise square root (Perronnin et al. 2010).”
Random Feature Maps for Dot Product Kernels (2012, background): “There have been approaches that try to perform embeddings in a task dependent manner (see for example Perronnin et al., 2010).”
Undoing the Damage of Dataset Bias (2012, methodology): “We evaluate our algorithms on cross-dataset generalization [14–16] in a holdone dataset out fashion.”
Improving the Fisher Kernel for Large-Scale Image Classification (2010, methodology): “[16] and Vedaldi and Zisserman [17] proposed different approximations for additive kernels.”
Large Scale Visual Recognition (2012, background): “Our work is also inspired by an emerging trend in computer vision studying large scale visual recognition [98, 36, 83, 30, 137, 52, 125, 20, 138], as well as fine-grained categorization [104, 139, 140, 141, 142].”
Towards good practice in large-scale learning for image classification (2012, methodology): “An explicit mapping of the image descriptors to efficiently deal with non-linear kernels was proposed in [21, 25, 36].”
Good Practice in Large-Scale Learning for Image Classification (2014, background): “Perronnin et al . later reﬁned this representation with a set of normalization steps [57].”
High-dimensional signature compression for large-scale image classification (2011, background): “[21] subsequently generalized this principle to any additive classifier.”
Action and Event Recognition with Fisher Vectors on a Compact Feature Set (2013, unknown): “The power normalization can be seen as an approximate explicit embedding of the χ(2) kernel [30].”
Three things everyone should know to improve object retrieval (2012, methodology): “This approach has been explored in the context of kernel maps for SVM classifiers by [22, 31].”
Efficient Classification for Additive Kernel SVMs (2013, background): “The more expensive table lookups can be avoided by rewriting the piecewise linear interpolation as a dot product of a dense vector of function values and a sparse vector indicating the bin indices, for e.g., see [34] or [45].”
Connecting the Dots with Landmarks: Discriminatively Learning Domain-Invariant Features for Unsupervised Domain Adaptation (2013, background): “For example, in computer vision, recent studies have shown that object classifiers optimized on one benchmark dataset often exhibit significant degradation in recognition accuracy when evaluated on another one (Torralba & Efros, 2011; Perronnin et al., 2010).”
Latent-Class Hough Forests for 3D Object Detection and Pose Estimation (2014, background): “In fact, it has been shown that significant performance degradation can occur when the negative training set is not representative of target domain [35,6,27].”
Author manuscript, published in "International Journal of Computer Vision (2013)" International Journal of Computer Vision manuscript No. (will be inserted by the editor) Image Classification with the Fisher Vector: Theory and Practice (2013, unknown): “Vedaldi and Zisserman (2010) and Perronnin et al (2010b) subsequently generalized this principle to any additive classiﬁer.”
Compositional Models for Video Event Detection: A Multiple Kernel Learning Latent Variable Approach (2013, background): “A body of work has aimed at providing efficient training and evaluation with kernelized classifiers via algorithmic optimizations or additive linear approximations [15, 6, 10].”
Geodesic flow kernel for unsupervised domain adaptation (2012, background): “Indeed, a stream of studies have shown that when image classifiers are evaluated outside of their training datasets, the performance degrades significantly [27, 9, 24].”
3D Pose from Motion for Cross-View Action Recognition via Non-linear Circulant Temporal Encoding (2014, unknown): “This result incorporates the finding that computing the square root of a BoW leads to better retrieval performance [17], and offers another perspective on the signed square root heuristic for BoW [9].”