Image categorization using Fisher kernels of non-iid image models

Download
2012-01-01
Cinbiş, Ramazan Gökberk
Schmid, Cordelia
The bag-of-words (BoW) model treats images as an unordered set of local regions and represents them by visual word histograms. Implicitly, regions are assumed to be identically and independently distributed (iid), which is a poor assumption from a modeling perspective. We introduce non-iid models by treating the parameters of BoW models as latent variables which are integrated out, rendering all local regions dependent. Using the Fisher kernel we encode an image by the gradient of the data log-likelihood w.r.t. hyper-parameters that control priors on the model parameters. Our representation naturally involves discounting transformations similar to taking square-roots, providing an explanation of why such transformations have proven successful. Using variational inference we extend the basic model to include Gaussian mixtures over local descriptors, and latent topic models to capture the co-occurrence structure of visual words, both improving performance. Our models yield state-of-the-art categorization performance using linear classifiers; without using non-linear transformations such as taking square-roots of features, or using (approximate) explicit embeddings of non-linear kernels.

Suggestions

Approximate Fisher Kernels of Non-iid Image Models for Image Categorization
Cinbiş, Ramazan Gökberk; Schmid, Cordelia (2016-06-01)
The bag-of-words (BoW) model treats images as sets of local descriptors and represents them by visual word histograms. The Fisher vector (FV) representation extends BoW, by considering the first and second order statistics of local descriptors. In both representations local descriptors are assumed to be identically and independently distributed (iid), which is a poor assumption from a modeling perspective. It has been experimentally observed that the performance of BoW and FV representations can be improved...
Image annotation with semi-supervised clustering
Sayar, Ahmet; Yarman Vural, Fatoş Tunay; Department of Computer Engineering (2009)
Image annotation is defined as generating a set of textual words for a given image, learning from the available training data consisting of visual image content and annotation words. Methods developed for image annotation usually make use of region clustering algorithms to quantize the visual information. Visual codebooks are generated from the region clusters of low level visual features. These codebooks are then, matched with the words of the text document related to the image, in various ways. In this th...
Texture segmentation using the mixtures of principal component analyzers
Musa, MEM; Duin, RPW; de Ridder, D; Atalay, Mehmet Volkan (2003-01-01)
The problem of segmenting an image into several modalities representing different textures can be modelled using Gaussian mixtures. Moreover, texture image patches when translated, rotated or scaled lie in low dimensional subspaces of the high-dimensional space spanned by the grey values. These two aspects make the mixture of local subspace models worth consideration for segmenting this type of images. In recent years a number of mixtures of local PCA models have been proposed. Most of these models require ...
FEATURE ENCODING MODELS FOR GEOGRAPHIC IMAGE RETRIEVAL AND CATEGORIZATION
Ozkan, Savas; Ates, Tayfun; Tola, Engin; Soysal, Medeni; Esen, Ersin (2014-04-25)
In this work, we survey the perormance of various feature encoding models for geographic image retrieval task Recently introduced Vector-of-Locally-Aggregated Descriptors (VLAD) and its Product Quantization encoded binary version VLAD-PQ are compared with the widely used Bag-of-Word (BoW) model. Evaluation results are shown on a publicly available 21-class LULC dataset. With experiments, it is shown that VLAD outperforms classical BoW representation albeit with some increases in the computation time. Additi...
Image Annotation With Semi-Supervised Clustering
Sayar, Ahmet; Yarman Vural, Fatoş Tunay (2009-09-16)
Methods developed for image annotation usually make use of region clustering algorithms. Visual codebooks are generated from the region clusters of low level features. These codebooks are then, matched with the words of the text document related to the image, in various ways. In this paper, we supervise the clustering process by using three types of side information. The first one is the topic probability information obtained from the text document associated with the image. The second is the orientation an...
Citation Formats
R. G. Cinbiş and C. Schmid, “Image categorization using Fisher kernels of non-iid image models,” presented at the 2012 IEEE Conference on Computer Vision and Pattern Recognition, Providence, RI, USA, 2012, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/56688.