Readspeaker menu

Novelty Detection

Is this a new object?

Paul Bodesheim, Alexander Freytag, Erik Rodner, Christoph Käding, and Joachim Denzler


Motivation

In many important learning tasks, training examples are often available for only one class. Learning in this scenario is difficult, since training examples from "outside" are not available at all. This problem is known as one-class classification (OCC), novelty detection, and outlier detection, to name just a few. Our work in this area addresses the use of methods for deriving a set of suitable OCC scores and to apply OCC in an incremental learning framework.


Null space methods

  Local Learning for Novelty Detection

[Bodesheim15:LND]

Paul Bodesheim and Alexander Freytag and Erik Rodner and Joachim Denzler. Local Novelty Detection in Multi-class Recognition Problems. IEEE Winter Conference on Applications of Computer Vision (WACV). 2015.   [pdf] [bib] [code] [github]
Short summary: In this paper, we propose using local learning for multi-class novelty detection, a framework that we call local novelty detection. Estimating the novelty of a new sample is an extremely challenging task due to the large variability of known object categories. The features used to judge on the novelty are often very specific for the object in the image and therefore we argue that individual novelty models for each test sample are important. Similar to human experts, it seems intuitive to first look for the most related images thus filtering out unrelated data. Afterwards, the system focuses on discovering similarities and differences to those images only. Therefore, we claim that it is beneficial to solely consider training images most similar to a test sample when deciding about its novelty. Following the principle of local learning, for each test sample a local novelty detection model is learned and evaluated. Our local novelty score turns out to be a valuable indicator for deciding whether the sample belongs to a known category from the training set or to a new, unseen one. With our local novelty detection approach, we achieve state-of-the-art performance in multi-class novelty detection on two popular visual object recognition datasets, Caltech-256 and ImageNet. We further show that our framework: (i) can be successfully applied to unknown face detection using the Labeled-Faces-in-the-Wild dataset and (ii) outperforms recent work on attribute-based unfamiliar class detection in fine-grained recognition of bird species on the challenging CUB-200-2011 dataset.




  nullspaceExample

[Bodesheim13:KNS]

Paul Bodesheim and Alexander Freytag and Erik Rodner and Michael Kemmler and Joachim Denzler: Kernel Null Space Methods for Novelty Detection. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2013. 3374--3381. [pdf] [bib] [project] [code] [github]
Short summary: Detecting samples from previously unknown classes is a crucial task in object recognition, especially when dealing with real-world applications where the closed-world assumption does not hold. We present how to apply a null space method for novelty detection, which maps all training samples of one class to a single point. Beside the possibility of modeling a single class, we are able to treat multiple known classes jointly and to detect novelties for a set of classes with a single model. In contrast to model the support of each known class individually, our approach makes use of a projection in a joint subspace where training samples of all known classes have zero intra-class variance. This subspace is called the null space of the training data. To decide about novelty of a test sample, our null space approach allows for solely relying on a distance measure instead of performing density estimation directly. Therefore, we derive a simple yet powerful method for multi-class novelty detection, an important problem not studied sufficiently so far. Our novelty detection approach is assessed in comprehensive multi-class experiments using the publicly available datasets Caltech-256 and ImageNet. The analysis reveals that our null space approach is perfectly suited for multi-class novelty detection since it outperforms all other approaches.

Entropy-based approaches

  overviewNovBasedOCC

[Bodesheim12:DOC]

Paul Bodesheim and Erik Rodner and Alexander Freytag and Joachim Denzler. Divergence-Based One-Class Classification Using Gaussian Processes. British Machine Vision Conference (BMVC). 2012. pp. 50.1--50-11. [pdf] [bib] [project] [poster] [extendedAbstract]
Short summary: We present an information theoretic framework for one-class classification, which allows for deriving several new novelty scores. With these scores, we are able to rank samples according to their novelty and to detect outliers not belonging to a learnt data distribution. The key idea of our approach is to measure the impact of a test sample on the previously learnt model. This is carried out in a probabilistic manner using Jensen-Shannon divergence and reclassification results derived from the Gaussian process regression framework. Our method is evaluated using well-known machine learning datasets as well as large-scale image categorisation experiments showing its ability to achieve state-of-the-art performance.

Approaches based on Gaussian processes

 

[Kemmler10:OCC]

Michael Kemmler and Erik Rodner and Joachim Denzler. One-Class Classification with Gaussian Processes. Asian Conference on Computer Vision (ACCV). 2010. 489--500. [pdf] [bib]

GPR_mean.jpgGPR_variance.jpg
   In many important task, training examples are often available for only one class. Learning the support of the category under focus is difficult, since training examples from "outside" are not available at all. This problem is known as one-class classification (OCC), novelty detection and outlier detection, to name just a few. The current paper addresses the use of Gaussian Processes for deriving a set of OCC scores. Empirical comparison with Support Vector Data Description (SVDD) on visual object recognition tasks using image kernels indicate the suitability of our approach.
dl2.png Code for one-class classification based on Gaussian process regression
 
 

[Rodner11:OCA]

Erik Rodner and Esther-Sabrina Wacker and Michael Kemmler and Joachim Denzler. One-Class Classification for Anomaly Detection in Wire Ropes with Gaussian Processes in a Few Lines of Code. IAPR Conference on Machine Vision Applications (MVA). 2011. 219--222. [pdf] [bib]

rope_example1.jpg
rope_example2.jpg
   Controlling ropes is an important task for ensuring the safety of elevators, bridges, or cable car transportation systems (see the page of Esther Wacker for a larger list of publications related to this topic). In practice, this is regularly done by a human observer who has to inspect every inch of a given rope searching for suspicious anomalies. An automatic procedure for solving this tedious task or for narrowing down the suspicious regions to a small fraction would be beneficial. In this approach, a recently proposed novelty detection method using Gaussian processes in combination with appearance-based HOG features is employed to tackle this problem. Experiments on a challenging ropeway dataset demonstrate that this approach outperforms existing state-of-the-art techniques such as Gaussian Mixture Models and Support Vector Data Description.
 
 

[Kemmler13:AIN]

Michael Kemmler and Erik Rodner and Petra Rösch and Jürgen Popp and Joachim Denzler. Automatic Identification of Novel Bacteria using Raman Spectroscopy and Gaussian Processes. Analytica Chimica Acta. 2013. (submitted). [pdf] [bib]
novelbac.jpg
   When classifying microorganisms, the typical closed-world assumption of pattern recognition does not hold. While today's databases are getting more and more complex, an exhaustive training database covering all possible species is not realistic due to the large biodiversity of microbes. In practice, it is therefore of utmost importance to detect novel microbes, which are not present during training. Our work analyzes the use of Gaussian process based one-class classification for this task. For a bacteria dataset comprising 50 different strains, we show its suitability in comparison to other state-of-the-art methods such as Gaussian mixture models, Parzen density estimation, and Support Vector Data Description.
 
 

[Bodesheim13:AOG]

Paul Bodesheim and Alexander Freytag and Erik Rodner and Joachim Denzler: Approximations of Gaussian Process Uncertainties for Visual Recognition Problems. Scandinavian Conference on Image Analysis (SCIA). 2013. pp. 182--194. [pdf] [bib] [project] [code]
approxGP-Var

 
   Gaussian processes offer the advantage of calculating the classification uncertainty in terms of predictive variance associated with the classification result. This is especially useful to select informative samples in active learning and to spot samples of previously unseen classes known as novelty detection. However, the Gaussian process framework suffers from high computational complexity leading to computation times too large for practical applications. Hence, we propose an approximation of the Gaussian process predictive variance leading to rigorous speedups. The complexity of both learning and testing the classification model regarding computational time and memory demand decreases by one order with respect to the number of training samples involved. The benefits of our approximations are verified in experimental evaluations for novelty detection and active learning of visual object categories on the datasets C-Pascal of Pascal VOC 2008, Caltech-256, and ImageNet.  
     

Publications

[Bodesheim15:LND]

Paul Bodesheim and Alexander Freytag and Erik Rodner and Joachim Denzler. Local Novelty Detection in Multi-class Recognition Problems. IEEE Winter Conference on Applications of Computer Vision (WACV). 2015.   [pdf] [bib] [code] [github]

[Bodesheim13:KNS]

Paul Bodesheim and Alexander Freytag and Erik Rodner and Michael Kemmler and Joachim Denzler: Kernel Null Space Methods for Novelty Detection. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2013. 3374--3381. [pdf] [bib] [project] [code] [github]

[Bodesheim13:AOG]

Paul Bodesheim and Alexander Freytag and Erik Rodner and Joachim Denzler: Approximations of Gaussian Process Uncertainties for Visual Recognition Problems. Scandinavian Conference on Image Analysis (SCIA). 2013. pp. 182--194. [pdf] [bib] [project] [code]

[Kemmler13:AIN]

Michael Kemmler and Erik Rodner and Petra Rösch and Jürgen Popp and Joachim Denzler. Automatic Identification of Novel Bacteria using Raman Spectroscopy and Gaussian Processes. Analytica Chimica Acta. 2013. (submitted). [pdf] [bib]

[Bodesheim12:DOC]

Paul Bodesheim and Erik Rodner and Alexander Freytag and Joachim Denzler. Divergence-Based One-Class Classification Using Gaussian Processes. British Machine Vision Conference (BMVC). 2012. pp. 50.1--50-11. [pdf] [bib] [project] [poster] [extendedAbstract]

[Rodner11:OCA]

Erik Rodner and Esther-Sabrina Wacker and Michael Kemmler and Joachim Denzler. One-Class Classification for Anomaly Detection in Wire Ropes with Gaussian Processes in a Few Lines of Code. IAPR Conference on Machine Vision Applications (MVA). 2011. 219--222. [pdf] [bib]

[Kemmler10:OCC]

Michael Kemmler and Erik Rodner and Joachim Denzler. One-Class Classification with Gaussian Processes. Asian Conference on Computer Vision (ACCV). 2010. 489--500. [pdf] [bib]