Incremental Linear Discriminant Analysis Using Sufficient Spanning Sets and Its Applications

This paper presents an incremental learning solution for Linear Discriminant Analysis (LDA) and its applications to object recognition problems. We apply the sufficient spanning set approximation in three steps i.e. update for the total scatter matrix, between-class scatter matrix and the projected data matrix, which leads an online solution which closely agrees with the batch solution in accuracy while significantly reducing the computational complexity. The algorithm yields an efficient solution to incremental LDA even when the number of classes as well as the set size is large. The incremental LDA method has been also shown useful for semi-supervised online learning. Label propagation is done by integrating the incremental LDA into an EM framework. The method has been demonstrated in the task of merging large datasets which were collected during MPEG standardization for face image retrieval, face authentication using the BANCA dataset, and object categorisation using the Caltech101 dataset. © 2010 Springer Science+Business Media, LLC.

Comparing crowding in human and ideal observers.

A visual target is more difficult to recognize when it is surrounded by other, similar objects. This breakdown in object recognition is known as crowding. Despite a long history of experimental work, computational models of crowding are still sparse. Specifically, few studies have examined crowding using an ideal-observer approach. Here, we compare crowding in ideal observers with crowding in humans. We derived an ideal-observer model for target identification under conditions of position and identity uncertainty. Simulations showed that this model reproduces the hallmark of crowding, namely a critical spacing that scales with viewing eccentricity. To examine how well the model fits quantitatively to human data, we performed three experiments. In Experiments 1 and 2, we measured observers' perceptual uncertainty about stimulus positions and identities, respectively, for a target in isolation. In Experiment 3, observers identified a target that was flanked by two distractors. We found that about half of the errors in Experiment 3 could be accounted for by the perceptual uncertainty measured in Experiments 1 and 2. The remainder of the errors could be accounted for by assuming that uncertainty (i.e., the width of internal noise distribution) about stimulus positions and identities depends on flanker proximity. Our results provide a mathematical restatement of the crowding problem and support the hypothesis that crowding behavior is a sign of optimality rather than a perceptual defect.

A neurophysiologically plausible population code model for feature integration explains visual crowding.

An object in the peripheral visual field is more difficult to recognize when surrounded by other objects. This phenomenon is called "crowding". Crowding places a fundamental constraint on human vision that limits performance on numerous tasks. It has been suggested that crowding results from spatial feature integration necessary for object recognition. However, in the absence of convincing models, this theory has remained controversial. Here, we present a quantitative and physiologically plausible model for spatial integration of orientation signals, based on the principles of population coding. Using simulations, we demonstrate that this model coherently accounts for fundamental properties of crowding, including critical spacing, "compulsory averaging", and a foveal-peripheral anisotropy. Moreover, we show that the model predicts increased responses to correlated visual stimuli. Altogether, these results suggest that crowding has little immediate bearing on object recognition but is a by-product of a general, elementary integration mechanism in early vision aimed at improving signal quality.

A Structured Model of Video Reproduces Primary Visual Cortical Organisation

The visual system must learn to infer the presence of objects and features in the world from the images it encounters, and as such it must, either implicitly or explicitly, model the way these elements interact to create the image. Do the response properties of cells in the mammalian visual system reflect this constraint? To address this question, we constructed a probabilistic model in which the identity and attributes of simple visual elements were represented explicitly and learnt the parameters of this model from unparsed, natural video sequences. After learning, the behaviour and grouping of variables in the probabilistic model corresponded closely to functional and anatomical properties of simple and complex cells in the primary visual cortex (V1). In particular, feature identity variables were activated in a way that resembled the activity of complex cells, while feature attribute variables responded much like simple cells. Furthermore, the grouping of the attributes within the model closely parallelled the reported anatomical grouping of simple cells in cat V1. Thus, this generative model makes explicit an interpretation of complex and simple cells as elements in the segmentation of a visual scene into basic independent features, along with a parametrisation of their moment-by-moment appearances. We speculate that such a segmentation may form the initial stage of a hierarchical system that progressively separates the identity and appearance of more articulated visual elements, culminating in view-invariant object recognition.

Single camera pose estimation using Bayesian filtering and Kinect motion priors

Traditional approaches to upper body pose estimation using monocular vision rely on complex body models and a large variety of geometric constraints. We argue that this is not ideal and somewhat inelegant as it results in large processing burdens, and instead attempt to incorporate these constraints through priors obtained directly from training data. A prior distribution covering the probability of a human pose occurring is used to incorporate likely human poses. This distribution is obtained offline, by fitting a Gaussian mixture model to a large dataset of recorded human body poses, tracked using a Kinect sensor. We combine this prior information with a random walk transition model to obtain an upper body model, suitable for use within a recursive Bayesian filtering framework. Our model can be viewed as a mixture of discrete Ornstein-Uhlenbeck processes, in that states behave as random walks, but drift towards a set of typically observed poses. This model is combined with measurements of the human head and hand positions, using recursive Bayesian estimation to incorporate temporal information. Measurements are obtained using face detection and a simple skin colour hand detector, trained using the detected face. The suggested model is designed with analytical tractability in mind and we show that the pose tracking can be Rao-Blackwellised using the mixture Kalman filter, allowing for computational efficiency while still incorporating bio-mechanical properties of the upper body. In addition, the use of the proposed upper body model allows reliable three-dimensional pose estimates to be obtained indirectly for a number of joints that are often difficult to detect using traditional object recognition strategies. Comparisons with Kinect sensor results and the state of the art in 2D pose estimation highlight the efficacy of the proposed approach.

Segmentation and recognition using structure from motion point clouds

We propose an algorithm for semantic segmentation based on 3D point clouds derived from ego-motion. We motivate five simple cues designed to model specific patterns of motion and 3D world structure that vary with object category. We introduce features that project the 3D cues back to the 2D image plane while modeling spatial layout and context. A randomized decision forest combines many such features to achieve a coherent 2D segmentation and recognize the object categories present. Our main contribution is to show how semantic segmentation is possible based solely on motion-derived 3D world structure. Our method works well on sparse, noisy point clouds, and unlike existing approaches, does not need appearance-based descriptors. Experiments were performed on a challenging new video database containing sequences filmed from a moving car in daylight and at dusk. The results confirm that indeed, accurate segmentation and recognition are possible using only motion and 3D world structure. Further, we show that the motion-derived information complements an existing state-of-the-art appearance-based method, improving both qualitative and quantitative performance. © 2008 Springer Berlin Heidelberg.

Testing of Depth-Encoded Hough Voting for Infrastructure Object Detection

The lack of viable methods to map and label existing infrastructure is one of the engineering grand challenges for the 21st century. For instance, over two thirds of the effort needed to geometrically model even simple infrastructure is spent on manually converting a cloud of points to a 3D model. The result is that few facilities today have a complete record of as-built information and that as-built models are not produced for the vast majority of new construction and retrofit projects. This leads to rework and design changes that can cost up to 10% of the installed costs. Automatically detecting building components could address this challenge. However, existing methods for detecting building components are not view and scale-invariant, or have only been validated in restricted scenarios that require a priori knowledge without considering occlusions. This leads to their constrained applicability in complex civil infrastructure scenes. In this paper, we test a pose-invariant method of labeling existing infrastructure. This method simultaneously detects objects and estimates their poses. It takes advantage of a recent novel formulation for object detection and customizes it to generic civil infrastructure scenes. Our preliminary experiments demonstrate that this method achieves convincing recognition results.

A performance evaluation of volumetric 3D interest point detectors

This paper presents the first performance evaluation of interest points on scalar volumetric data. Such data encodes 3D shape, a fundamental property of objects. The use of another such property, texture (i.e. 2D surface colouration), or appearance, for object detection, recognition and registration has been well studied; 3D shape less so. However, the increasing prevalence of 3D shape acquisition techniques and the diminishing returns to be had from appearance alone have seen a surge in 3D shape-based methods. In this work, we investigate the performance of several state of the art interest points detectors in volumetric data, in terms of repeatability, number and nature of interest points. Such methods form the first step in many shape-based applications. Our detailed comparison, with both quantitative and qualitative measures on synthetic and real 3D data, both point-based and volumetric, aids readers in selecting a method suitable for their application. © 2012 Springer Science+Business Media, LLC.