Is a Detector Only Good for Detection?

A common design of an object recognition system has two steps, a detection step followed by a foreground within-class classification step. For example, consider face detection by a boosted cascade of detectors followed by face ID recognition via one-vs-all (OVA) classifiers. Another example is human detection followed by pose recognition. Although the detection step can be quite fast, the foreground within-class classification process can be slow and becomes a bottleneck. In this work, we formulate a filter-and-refine scheme, where the binary outputs of the weak classifiers in a boosted detector are used to identify a small number of candidate foreground state hypotheses quickly via Hamming distance or weighted Hamming distance. The approach is evaluated in three applications: face recognition on the FRGC V2 data set, hand shape detection and parameter estimation on a hand data set and vehicle detection and view angle estimation on a multi-view vehicle data set. On all data sets, our approach has comparable accuracy and is at least five times faster than the brute force approach.

Non-Rigid Shape from Image Streams

We present a framework for estimating 3D relative structure (shape) and motion given objects undergoing nonrigid deformation as observed from a fixed camera, under perspective projection. Deforming surfaces are approximated as piece-wise planar, and piece-wise rigid. Robust registration methods allow tracking of corresponding image patches from view to view and recovery of 3D shape despite occlusions, discontinuities, and varying illumination conditions. Many relatively small planar/rigid image patch trackers are scattered throughout the image; resulting estimates of structure and motion at each patch are combined over local neighborhoods via an oriented particle systems formulation. Preliminary experiments have been conducted on real image sequences of deforming objects and on synthetic sequences where ground truth is known.

Investigating the causal role of rightOFA in holistic detection of Mooney faces and objects: an fMRI-guided TMS study

Background The right occipital face area (rOFA) is known to be involved in face discrimination based on local featural information. Whether this region is involved in global, holistic stimulus processing is not known. Objective We used fMRI-guided transcranial magnetic stimulation (TMS) to investigate whether rOFA is causally implicated in stimulus detection based on holistic processing, by the use of Mooney stimuli. Methods Two studies were carried out: In Experiment 1, participants performed a detection task involving Mooney faces and Mooney objects; Mooney stimuli lack distinguishable local features and can be detected solely via holistic processing (i.e. at a global level) with top-down guidance from previously stored representations. Experiment 2 required participants to detect shapes which are recognized via bottom-up integration of local (collinear) Gabor elements and was performed to control for specificity of rightOFA's implication in holistic detection. Results In Experiment 1, TMS over rOFA and rLO impaired detection of all stimulus categories, with no category-specific effect. In Experiment 2, shape detection was impaired when TMS was applied over rLO but not over rOFA. Conclusions Our results demonstrate that rOFA is causally implicated in the type of top-down holistic detection required by Mooney stimuli and that such role is not face-selective. In contrast, rOFA does not appear to play a causal role in in detection of shapes based on bottom-up integration of local components, demonstrating that its involvement in processing non-face stimuli is specific for holistic processing.

L’impact de la stéréoscopie dans la reconnaissance, la perception et la constance de forme 3D

Les buts des recherches présentées dans cette thèse étaient d’évaluer le rôle de la stéréoscopie dans la reconnaissance de forme, dans la perception du relief et dans la constance de forme. La première étude a examiné le rôle de la stéréoscopie dans la perception des formes visuelles en utilisant une tâche de reconnaissance de formes. Les stimuli pouvaient être présentés en 2D, avec disparité normale (3D) ou avec disparité inversée. La performance de reconnaissance était meilleure avec les modes de présentation 2D et 3D qu’avec la 3D inversée. Cela indique que la stéréoscopie contribue à la reconnaissance de forme. La deuxième étude s’est intéressée à la contribution conjointe de l’ombrage et de la stéréoscopie dans la perception du relief des formes. Les stimuli étaient des images d’une forme 3D convexe synthétique présentée sous un point de vue menant à une ambigüité quant à sa convexité. L’illumination pouvait provenir du haut ou du bas et de la gauche ou de la droite, et les stimuli étaient présentés dichoptiquement avec soit de la disparité binoculaire normale, de la disparité inversée ou sans disparité entre les vues. Les participants ont répondu que les formes étaient convexes plus souvent lorsque la lumière provenait du haut que du bas, plus souvent avec la disparité normale qu’en 2D, et plus souvent avec absence de disparité qu’avec disparité inversée. Les effets de direction d’illumination et du mode de présentation étaient additifs, c’est-à-dire qu’ils n’interagissaient pas. Cela indique que l’ombrage et la stéréoscopie contribuent indépendamment à la perception du relief des formes. La troisième étude a évalué la contribution de la stéréoscopie à la constance de forme, et son interaction avec l’expertise perceptuelle. Elle a utilisé trois tâches de discrimination séquentielle de trombones tordus ayant subi des rotations en profondeur. Les stimuli pouvaient être présentés sans stéréoscopie, avec stéréoscopie normale ou avec stéréoscopie inversée. Dans la première moitié de l’Exp. 1, dans laquelle les variations du mode de présentation étaient intra-sujets, les performances étaient meilleures en 3D qu’en 2D et qu’en 3D inversée. Ces effets ont été renversés dans la seconde moitié de l’expérience, et les coûts de rotation sont devenus plus faibles pour la 2D et la 3D inversée que pour la 3D. Dans les Exps. 2 (variations intra-sujets du mode de présentation, avec un changement de stimuli au milieu de l’expérience) et 3 (variations inter-sujets du mode de présentation), les effets de rotation étaient en tout temps plus faibles avec stéréoscopie qu’avec stéréoscopie inversée et qu’en 2D, et plus faibles avec stéréoscopie inversée que sans stéréoscopie. Ces résultats indiquent que la stéréoscopie contribue à la constance de forme. Toutefois, cela demande qu’elle soit valide avec un niveau minimal de consistance, sinon elle devient stratégiquement ignorée. En bref, les trois études présentées dans cette thèse ont permis de montrer que la stéréoscopie contribue à la reconnaissance de forme, à la perception du relief et à la constance de forme. De plus, l’ombrage et la stéréoscopie sont intégrés linéairement.

Personalized 3D reconstruction of the rib cage for clinical assessment of trunk deformities

Scoliosis is a 3D deformity of the spine and rib cage. Extensive validation of 3D reconstruction methods of the spine from biplanar radiography has already been published. In this article, we propose a novel method to reconstruct the rib cage, using the same biplanar views as for the 3D reconstruction of the spine, to allow clinical assessment of whole trunk deformities. This technique uses a semi-automatic segmentation of the ribs in the postero-anterior X-ray view and an interactive segmentation of partial rib edges in the lateral view. The rib midlines are automatically extracted in 2D and reconstructed in 3D using the epipolar geometry. For the ribs not visible in the lateral view, the method predicts their 3D shape. The accuracy of the proposed method has been assessed using data obtained from a synthetic bone model as a gold standard and has also been evaluated using data of real patients with scoliotic deformities. Results show that the reconstructed ribs enable a reliable evaluation of the rib axial rotation, which will allow a 3D clinical assessment of the spine and rib cage deformities.

Combining 3D and 2D for less constrained periocular recognition

Periocular recognition has recently become an active topic in biometrics. Typically it uses 2D image data of the periocular region. This paper is the first description of combining 3D shape structure with 2D texture. A simple and effective technique using iterative closest point (ICP) was applied for 3D periocular region matching. It proved its strength for relatively unconstrained eye region capture, and does not require any training. Local binary patterns (LBP) were applied for 2D image based periocular matching. The two modalities were combined at the score-level. This approach was evaluated using the Bosphorus 3D face database, which contains large variations in facial expressions, head poses and occlusions. The rank-1 accuracy achieved from the 3D data (80%) was better than that for 2D (58%), and the best accuracy (83%) was achieved by fusing the two types of data. This suggests that significant improvements to periocular recognition systems could be achieved using the 3D structure information that is now available from small and inexpensive sensors.

Reverse correlation reveals how observers sample visual information when estimating three-dimensional shape

Human observers exhibit large systematic distance-dependent biases when estimating the three-dimensional (3D) shape of objects defined by binocular image disparities. This has led some to question the utility of disparity as a cue to 3D shape and whether accurate estimation of 3D shape is at all possible. Others have argued that accurate perception is possible, but only with large continuous perspective transformations of an object. Using a stimulus that is known to elicit large distance-dependent perceptual bias (random dot stereograms of elliptical cylinders) we show that contrary to these findings the simple adoption of a more naturalistic viewing angle completely eliminates this bias. Using behavioural psychophysics, coupled with a novel surface-based reverse correlation methodology, we show that it is binocular edge and contour information that allows for accurate and precise perception and that observers actively exploit and sample this information when it is available.

Individuazione di punti salienti in dati 3D mediante rappresentazioni strutturate

Questa tesi si inserisce nel filone di ricerca dell'elaborazione di dati 3D, e in particolare nella 3D Object Recognition, e delinea in primo luogo una panoramica sulle principali rappresentazioni strutturate di dati 3D, le quali rappresentano una prerogativa necessaria per implementare in modo efficiente algoritmi di processing di dati 3D, per poi presentare un nuovo algoritmo di 3D Keypoint Detection che è stato sviluppato e proposto dal Computer Vision Laboratory dell'Università di Bologna presso il quale ho effettuato la mia attività di tesi.

Statistical shape analysis via principal factor analysis

Statistical shape analysis techniques commonly employed in the medical imaging community, such as active shape models or active appearance models, rely on principal component analysis (PCA) to decompose shape variability into a reduced set of interpretable components. In this paper we propose principal factor analysis (PFA) as an alternative and complementary tool to PCA providing a decomposition into modes of variation that can be more easily interpretable, while still being a linear efficient technique that performs dimensionality reduction (as opposed to independent component analysis, ICA). The key difference between PFA and PCA is that PFA models covariance between variables, rather than the total variance in the data. The added value of PFA is illustrated on 2D landmark data of corpora callosa outlines. Then, a study of the 3D shape variability of the human left femur is performed. Finally, we report results on vector-valued 3D deformation fields resulting from non-rigid registration of ventricles in MRI of the brain.

THE THREE DIMENSIONAL SHAPE AND ROUGHNESS OF MINERAL DUST

Mineral dust shape and roughness are important for a multitude of processes; it is known for aspherical shape but the true measurements in three dimensions are rare. Atomic Force Microscope was used for determine both 3D shape and roughness for two dust which are commonly used in laboratory experiments – Arizona Test Dust (ATD) and Kaolinite. We determined both of them are rather flat and round; an oblate spheroid would be a good model. Loess Filter was used to smooth the particles' surface and correlation analysis was used to examine the surfaces' properties of the dust; we found no features under 100nm scales. Also, our particles' surface area result is very similar to BET surface area.

Comparison of 3D kinetic and hydrodynamic models to ROSINA-COPS measurements of the neutral coma of 67P/Churyumov-Gerasimenko

67P/Churyumov-Gerasimenko (67P) is a Jupiter-family comet and the object of investigation of the European Space Agency mission Rosetta. This report presents the first full 3D simulation results of 67P’s neutral gas coma. In this study we include results from a direct simulation Monte Carlo method, a hydrodynamic code, and a purely geometric calculation which computes the total illuminated surface area on the nucleus. All models include the triangulated 3D shape model of 67P as well as realistic illumination and shadowing conditions. The basic concept is the assumption that these illumination conditions on the nucleus are the main driver for the gas activity of the comet. As a consequence, the total production rate of 67P varies as a function of solar insolation. The best agreement between the model and the data is achieved when gas fluxes on the night side are in the range of 7% to 10% of the maximum flux, accounting for contributions from the most volatile components. To validate the output of our numerical simulations we compare the results of all three models to in situ gas number density measurements from the ROSINA COPS instrument. We are able to reproduce the overall features of these local neutral number density measurements of ROSINA COPS for the time period between early August 2014 and January 1 2015 with all three models. Some details in the measurements are not reproduced and warrant further investigation and refinement of the models. However, the overall assumption that illumination conditions on the nucleus are at least an important driver of the gas activity is validated by the models. According to our simulation results we find the total production rate of 67P to be constant between August and November 2014 with a value of about 1 × 10²⁶ molecules s⁻¹.

PDB-Explorer: a web-based interactive map of the protein data bank in shape space

Background The RCSB Protein Data Bank (PDB) provides public access to experimentally determined 3D-structures of biological macromolecules (proteins, peptides and nucleic acids). While various tools are available to explore the PDB, options to access the global structural diversity of the entire PDB and to perceive relationships between PDB structures remain very limited. Methods A 136-dimensional atom pair 3D-fingerprint for proteins (3DP) counting categorized atom pairs at increasing through-space distances was designed to represent the molecular shape of PDB-entries. Nearest neighbor searches examples were reported exemplifying the ability of 3DP-similarity to identify closely related biomolecules from small peptides to enzyme and large multiprotein complexes such as virus particles. The principle component analysis was used to obtain the visualization of PDB in 3DP-space. Results The 3DP property space groups proteins and protein assemblies according to their 3D-shape similarity, yet shows exquisite ability to distinguish between closely related structures. An interactive website called PDB-Explorer is presented featuring a color-coded interactive map of PDB in 3DP-space. Each pixel of the map contains one or more PDB-entries which are directly visualized as ribbon diagrams when the pixel is selected. The PDB-Explorer website allows performing 3DP-nearest neighbor searches of any PDB-entry or of any structure uploaded as protein-type PDB file. All functionalities on the website are implemented in JavaScript in a platform-independent manner and draw data from a server that is updated daily with the latest PDB additions, ensuring complete and up-to-date coverage. The essentially instantaneous 3DP-similarity search with the PDB-Explorer provides results comparable to those of much slower 3D-alignment algorithms, and automatically clusters proteins from the same superfamilies in tight groups. Conclusion A chemical space classification of PDB based on molecular shape was obtained using a new atom-pair 3D-fingerprint for proteins and implemented in a web-based database exploration tool comprising an interactive color-coded map of the PDB chemical space and a nearest neighbor search tool. The PDB-Explorer website is freely available at www.​cheminfo.​org/​pdbexplorer and represents an unprecedented opportunity to interactively visualize and explore the structural diversity of the PDB.

Optimization of TRPV6 Calcium Channel Inhibitors Using a 3D Ligand-Based Virtual Screening Method

Herein, we report the discovery of the first potent and selective inhibitor of TRPV6, a calcium channel overexpressed in breast and prostate cancer, and its use to test the effect of blocking TRPV6-mediated Ca2+-influx on cell growth. The inhibitor was discovered through a computational method, xLOS, a 3D-shape and pharmacophore similarity algorithm, a type of ligand-based virtual screening (LBVS) method described briefly here. Starting with a single weakly active seed molecule, two successive rounds of LBVS followed by optimization by chemical synthesis led to a selective molecule with 0.3 μM inhibition of TRPV6. The ability of xLOS to identify different scaffolds early in LBVS was essential to success. The xLOS method may be generally useful to develop tool compounds for poorly characterized targets.

Shape from photographs:a multi-view stereo pipeline

Acquiring 3D shape from images is a classic problem in Computer Vision occupying researchers for at least 20 years. Only recently however have these ideas matured enough to provide highly accurate results. We present a complete algorithm to reconstruct 3D objects from images using the stereo correspondence cue. The technique can be described as a pipeline of four basic building blocks: camera calibration, image segmentation, photo-consistency estimation from images, and surface extraction from photo-consistency. In this Chapter we will put more emphasis on the latter two: namely how to extract geometric information from a set of photographs without explicit camera visibility, and how to combine different geometry estimates in an optimal way. © 2010 Springer-Verlag Berlin Heidelberg.

Reconstruction of surfaces of revolution from single uncalibrated views

This paper addresses the problem of recovering the 3D shape of a surface of revolution from a single uncalibrated perspective view. The algorithm introduced here makes use of the invariant properties of a surface of revolution and its silhouette to locate the image of the revolution axis, and to calibrate the focal length of the camera. The image is then normalized and rectified such that the resulting silhouette exhibits bilateral symmetry. Such a rectification leads to a simpler differential analysis of the silhouette, and yields a simple equation for depth recovery. It is shown that under a general camera configuration, there will be a 2-parameter family of solutions for the reconstruction. The first parameter corresponds to an unknown scale, whereas the second one corresponds to an unknown attitude of the object. By identifying the image of a latitude circle, the ambiguity due to the unknown attitude can be resolved. Experimental results on real images are presented, which demonstrate the quality of the reconstruction. © 2004 Elsevier B.V. All rights reserved.