Eric O. Postma

Image processing for artist identification

A survey of the literature reveals that image processing tools aimed at supplementing the art historians toolbox are currently in the earliest stages of development. To jump-start the development of such methods, the Van Gogh and Kroller-Muller museums in The Netherlands agreed to make a data set of 101 high-resolution gray-scale scans of paintings within their collections available to groups of image processing researchers from several different universities. This article describes the approaches to brushwork analysis and artist identification developed by three research groups, within the framework of this data set.

Development and external validation of a predictive model for pathological complete response of rectal cancer patients including sequential PET-CT imaging

PURPOSEnTo develop and validate an accurate predictive model and a nomogram for pathologic complete response (pCR) after chemoradiotherapy (CRT) for rectal cancer based on clinical and sequential PET-CT data. Accurate prediction could enable more individualised surgical approaches, including less extensive resection or even a wait-and-see policy.nnnMETHODS AND MATERIALSnPopulation based databases from 953 patients were collected from four different institutes and divided into three groups: clinical factors (training: 677 patients, validation: 85 patients), pre-CRT PET-CT (training: 114 patients, validation: 37 patients) and post-CRT PET-CT (training: 107 patients, validation: 55 patients). A pCR was defined as ypT0N0 reported by pathology after surgery. The data were analysed using a linear multivariate classification model (support vector machine), and the models performance was evaluated using the area under the curve (AUC) of the receiver operating characteristic (ROC) curve.nnnRESULTSnThe occurrence rate of pCR in the datasets was between 15% and 31%. The model based on clinical variables (AUC(train)=0.61±0.03, AUC(validation)=0.69±0.08) resulted in the following predictors: cT- and cN-stage and tumour length. Addition of pre-CRT PET data did not result in a significantly higher performance (AUC(train)=0.68±0.08, AUC(validation)=0.68±0.10) and revealed maximal radioactive isotope uptake (SUV(max)) and tumour location as extra predictors. The best model achieved was based on the addition of post-CRT PET-data (AUC(train)=0.83±0.05, AUC(validation)=0.86±0.05) and included the following predictors: tumour length, post-CRT SUV(max) and relative change of SUV(max). This model performed significantly better than the clinical model (p(train)<0.001, p(validation)=0.056).nnnCONCLUSIONSnThe model and the nomogram developed based on clinical and sequential PET-CT data can accurately predict pCR, and can be used as a decision support tool for surgery after prospective validation.

SCAN: a scalable model of attentional selection

This paper describes the SCAN (Signal Channelling Attentional Network) model, a scalable neural network model for attentional scanning. The building block of SCAN is a gating lattice, a sparsely-connected neural network defined as a special case of the Ising lattice from statistical mechanics. The process of spatial selection through covert attention is interpreted as a biological solution to the problem of translation-invariant pattern processing. In SCAN, a sequence of pattern translations combines active selection with translation-invariant processing. Selected patterns are channelled through a gating network, formed by a hierarchical fractal structure of gating lattices, and mapped onto an output window. We show how the incorporation of an expectation-generating classifier network (e.g. Carpenter and Grossbergs ART network) into SCAN allows attentional selection to be driven by expectation. Simulation studies show the SCAN model to be capable of attending and identifying object patterns that are part of a realistically sized natural image. Copyright 1997 Elsevier Science Ltd.

PURSUING AUTOMATED CLASSIFICATION OF HISTORIC PHOTOGRAPHIC PAPERS FROM RAKING LIGHT IMAGES

Abstract Surface texture is a critical feature in the manufacture, marketing, and use of photographic paper. Raking light reveals texture through a stark rendering of highlights and shadows. Though close-up raking light images effectively document surface features of photographic paper, the sheer number and diversity of textures used for historic papers prohibits efficient visual classification. This work provides evidence that automatic, computer-based classification of texture documented with raking light is feasible by demonstrating an encouraging degree of success sorting a set of 120 images made from samples of historic silver gelatin paper. Using this dataset, four university teams applied different image-processing strategies for automatic feature extraction and degree of similarity quantification. All four approaches successfully detected strong affinities and outliers built into the dataset. The creation and deployment of the algorithms was carried out by the teams without prior knowledge of the distributions of similarities and outliers. These results indicate that automatic classification of silver gelatin photographic paper based on close-up texture images is feasible and should be pursued. To encourage the development of other classification schemes, the 120-sample “training” dataset used in this work is available to other academic researchers at http://www.PaperTextureID.org.

Reactive agents and perceptual ambiguity

Reactive agents are generally believed to be incapable of coping with perceptual ambiguity (i.e., identical sensory states that require different responses). However, a recent finding suggests that reactive agents can cope with perceptual ambiguity in a simple model (Nolfi, 2002). This paper investigates to what extent reactive and nonreactive agents can cope with perceptual ambiguity, and which strategies are employed when doing so. A model of active categorical perception (called Acp) is introduced. In Acp, situated agents with different types of neurocontrollers are optimized to categorize objects by adaptively coordinating action and perception. Our experiments show that both nonreactive and reactive agents can cope with perceptual ambiguity. An analysis of the behavior reveals that nonreactive agents use their internal memory to cope with perceptual ambiguity, while reactive agents use the environment as an external memory to compensate for their lack of an internal memory. We conclude that reactive agents can cope with perceptual ambiguity in the context of active categorical perception, and that they can organize their behavior according to stimuli that are no longer present, especially when they incorporate a nonlinear sensorimotor mapping. Moreover, we may conclude that sensory state-transition diagrams provide insight into the strategies employed by reactive agents to deal with perceptual ambiguity, and their use of the environment as an external memory.

Toward Discovery of the Artist?s Style: Learning to recognize artists by their artworks

Author attribution through the recognition of visual characteristics is a commonly used approach by art experts. By studying a vast number of artworks, art experts acquire the ability to recognize the unique characteristics of artists. In this article, we present an approach that uses the same principles to discover the characteristic features that determine an artists touch. By training a convolutional neural network (PigeoNET) on a large collection of digitized artworks to perform the task of automatic artist attribution, the network is encouraged to discover artist-specific visual features. The trained network is shown to be capable of attributing previously unseen artworks to the actual artists with an accuracy of more than 70%. In addition, the trained network provides fine-grained information about the artist-specific characteristics of spatial regions within the artworks. We demonstrate this ability by means of a single artwork that combines characteristics of two closely collaborating artists. PigeoNET generates a visualization that indicates for each location on the artwork who is the most likely artist to have contributed to the visual characteristics at that location. We conclude that PigeoNET represents a fruitful approach for the future of computer-supported examination of artworks.

Learning scale-variant and scale-invariant features for deep image classification

Abstract Convolutional Neural Networks (CNNs) require large image corpora to be trained on classification tasks. The variation in image resolutions, sizes of objects and patterns depicted, and image scales, hampers CNN training and performance, because the task-relevant information varies over spatial scales. Previous work attempting to deal with such scale variations focused on encouraging scale-invariant CNN representations. However, scale-invariant representations are incomplete representations of images, because images contain scale-variant information as well. This paper addresses the combined development of scale-invariant and scale-variant representations. We propose a multi-scale CNN method to encourage the recognition of both types of features and evaluate it on a challenging image classification task involving task-relevant characteristics at multiple scales. The results show that our multi-scale CNN outperforms single-scale CNN. This leads to the conclusion that encouraging the combined development of a scale-invariant and scale-variant representation in CNNs is beneficial to image recognition performance.

Outlier Detection with One-Class Classifiers from ML and KDD

The problem of outlier detection is well studied in the fields of Machine Learning (ML) and Knowledge Discovery in Databases (KDD). Both fields have their own methods and evaluation procedures. In ML, Support Vector Machines and Parzen Windows are well-known methods that can be used for outlier detection. In KDD, the heuristic local-density estimation methods LOF and LOCI are generally considered to be superior outlier-detection methods. Hitherto, the performances of these ML and KDD methods have not been compared. This paper formalizes LOF and LOCI in the ML framework of one-class classification and performs a comparative evaluation of the ML and KDD outlier-detection methods on real-world datasets. Experimental results show that LOF and SVDD are the two best-performing methods. It is concluded that both fields offer outlier-detection methods that are competitive in performance and that bridging the gap between both fields may facilitate the development of outlier-detection methods.

Adaptive gaze control for object detection

We propose a novel gaze-control model for detecting objects in images. The model, named act-detect, uses the information from local image samples in order to shift its gaze towards object locations. The model constitutes two main contributions. The first contribution is that the model’s setup makes it computationally highly efficient in comparison with existing window-sliding methods for object detection, while retaining an acceptable detection performance. act-detect is evaluated on a face-detection task using a publicly available image set. In terms of detection performance, act-detect slightly outperforms the window-sliding methods that have been applied to the face-detection task. In terms of computational efficiency, act-detect clearly outperforms the window-sliding methods: it requires in the order of hundreds fewer samples for detection. The second contribution of the model lies in its more extensive use of local samples than previous models: instead of merely using them for verifying object presence at the gaze location, the model uses them to determine a direction and distance to the object of interest. The simultaneous adaptation of both the model’s visual features and its gaze-control strategy leads to the discovery of features and strategies for exploiting the local context of objects. For example, the model uses the spatial relations between the bodies of the persons in the images and their faces. The resulting gaze control is a temporal process, in which the object’s context is exploited at different scales and at different image locations relative to the object.

Macroscopic analysis of robot foraging behaviour

Microscopic analysis is a standard approach in the study of robot behaviour. Typically, the approach comprises the analysis of a single (or sometimes a few) robot–environment system(s) to reveal specific properties of robot behaviour. In contrast to microscopic analysis, macroscopic analysis focuses on averaged properties of systems. The advantage is that such a property is easier to generalize so that it can be established to what extent the property is universal. This paper investigates whether a macroscopic analysis can reveal a universal property of adaptive behaviour in a robot model of foraging behaviour. Our analysis reveals that the step lengths of the most successful robots are distributed according to a Lévy-flight distribution. From studies on a variety of natural species, it is known that such a distribution constitutes a universal property of foraging behaviour. Thereafter, we discuss an example of how macroscopic analysis can be applied to existing research in evolutionary robotics, and relate the macroscopic and microscopic analyses of foraging behaviour to the framework of scientific research described by Cohen (1995, Empirical Methods for Artificial Intelligence (Cambridge MA: MIT Press)). We conclude that macroscopic analysis may predict universal properties of adaptive behaviour and that it may complement microscopic analysis in the study of adaptive behaviour.