An adaptation framework for head-pose classification in dynamic multi-view scenarios

Multi-view head-pose estimation in low-resolution, dynamic scenes is difficult due to blurred facial appearance and perspective changes as targets move around freely in the environment. Under these conditions, acquiring sufficient training examples to learn the dynamic relationship between position, face appearance and head-pose can be very expensive. Instead, a transfer learning approach is proposed in this work. Upon learning a weighted-distance function from many examples where the target position is fixed, we adapt these weights to the scenario where target positions are varying. The adaptation framework incorporates reliability of the different face regions for pose estimation under positional variation, by transforming the target appearance to a canonical appearance corresponding to a reference scene location. Experimental results confirm effectiveness of the proposed approach, which outperforms state-of-the-art by 9.5% under relevant conditions. To aid further research on this topic, we also make DPOSE- a dynamic, multi-view head-pose dataset with ground-truth publicly available with this paper.

A unified view of energetic efficiency in active drag reduction, thrust generation and self-propulsion through a loss coefficient with some applications

An analysis of the energy budget for the general case of a body translating in a stationary fluid under the action of an external force is used to define a power loss coefficient. This universal definition of power loss coefficient gives a measure of the energy lost in the wake of the translating body and, in general, is applicable to a variety of flow configurations including active drag reduction, self-propulsion and thrust generation. The utility of the power loss coefficient is demonstrated on a model bluff body flow problem concerning a two-dimensional elliptical cylinder in a uniform cross-flow. The upper and lower boundaries of the elliptic cylinder undergo continuous motion due to a prescribed reflectionally symmetric constant tangential surface velocity. It is shown that a decrease in drag resulting from an increase in the strength of tangential surface velocity leads to an initial reduction and eventual rise in the power loss coefficient. A maximum in energetic efficiency is attained for a drag reducing tangential surface velocity which minimizes the power loss coefficient. The effect of the tangential surface velocity on drag reduction and self-propulsion of both bluff and streamlined bodies is explored through a variation in the thickness ratio (ratio of the minor and major axes) of the elliptical cylinders.

Exploring Transfer Learning Approaches for Head Pose Classification from Multi-view Surveillance Images

Head pose classification from surveillance images acquired with distant, large field-of-view cameras is difficult as faces are captured at low-resolution and have a blurred appearance. Domain adaptation approaches are useful for transferring knowledge from the training (source) to the test (target) data when they have different attributes, minimizing target data labeling efforts in the process. This paper examines the use of transfer learning for efficient multi-view head pose classification with minimal target training data under three challenging situations: (i) where the range of head poses in the source and target images is different, (ii) where source images capture a stationary person while target images capture a moving person whose facial appearance varies under motion due to changing perspective, scale and (iii) a combination of (i) and (ii). On the whole, the presented methods represent novel transfer learning solutions employed in the context of multi-view head pose classification. We demonstrate that the proposed solutions considerably outperform the state-of-the-art through extensive experimental validation. Finally, the DPOSE dataset compiled for benchmarking head pose classification performance with moving persons, and to aid behavioral understanding applications is presented in this work.

VIRTUAL VIEW SYNTHESIS BY NON-LOCAL MEANS FILTERING USING TEMPORAL DATA

A novel algorithm for Virtual View Synthesis based on Non-Local Means Filtering is presented in this paper. Apart from using the video frames from the nearby cameras and the corresponding per-pixel depth map, this algorithm also makes use of the previously synthesized frame. Simple and efficient, the algorithm can synthesize video at any given virtual viewpoint at a faster rate. In the process, the quality of the synthesized frame is not compromised. Experimental results prove the above mentioned claim. The subjective and objective quality of the synthesized frames are comparable to the existing algorithms.

A Suzaku view of IGR J16393-4643

The pulsar IGR J16393-4643 belongs to a class of highly absorbed supergiant high-mass X-ray binaries (HMXBs), characterized by a very high column density of absorbing matter. We present the results of simultaneous broad-band pulsation and spectrum analysis from a 44-ks Suzaku observation of the source. The orbital intensity profile created with the Swift Burst Alert Telescope (Swift-BAT) light curve shows an indication of IGR J16393-4643 being an eclipsing system with a short eclipse semi-angle theta(E) similar to 17 degrees. For a supergiant companion star with a 20-R-circle dot radius, this implies an inclination of the orbital plane in the range 39 degrees-57 degrees, whereas for a main-sequence B star as the companion with a 10-R-circle dot radius, the inclination of the orbital plane is in the range 60 degrees-77 degrees. Pulse profiles created for different energy bands have complex morphology, which shows some energy dependence and increases in pulse fraction with energy. We have also investigated broad-band spectral characteristics, phase-averaged spectra and resolving the pulse phase into peak and trough phases. The phase-averaged spectrum has a very high N-H(similar to 3 x 10(23) cm(-2)) and is described by a power law (Gamma similar to 0.9) with a high-energy cut-off above 20 keV. We find a change in the spectral index in the peak and trough phases, implying an underlying change in the source spectrum.

Lizards of the Thar Desert - Resource partitioning and community composition

How similar species co-exist in nature is a fundamental question in community ecology. Resource partitioning has been studied in desert lizard communities across four continents, but data from South Asia is lacking. We used area-constrained visual encounter surveys to study community composition and spatial and temporal resource partitioning in a lizard community during summer in the Thar Desert, western India, addressing an important biogeographic gap in knowledge. Twelve one-hectare grids divided into 25 m x 25 m plots were placed across four habitats barren dunes, stabilized dunes, grassland, and rocky hills. We recorded 1039 sightings of 12 species during 84 sampling sessions. Lizard abundance decreased in the order stabilized dunes > grassland > barren dunes > rocky hills; richness was in roughly the opposite order. Resource partitioning was examined for the seven commonest species. Overall spatial overlap was low (<0.6) between species pairs. Overlap was higher within habitats, but species showed finer separation through use of different microhabitat categories and specific spatial resources, as well as by positioning at different distances to vegetation. Diurnal species were also separated by peak time of activity. Space appears to be an important resource dimension facilitating coexistence in this desert lizard community. (C) 2015 Elsevier Ltd. All rights reserved.

Dynamics of 3D view invariance in monkey inferotemporal cortex

Rotations in depth are challenging for object vision because features can appear, disappear, be stretched or compressed. Yet we easily recognize objects across views. Are the underlying representations view invariant or dependent? This question has been intensely debated in human vision, but the neuronal representations remain poorly understood. Here, we show that for naturalistic objects, neurons in the monkey inferotemporal (IT) cortex undergo a dynamic transition in time, whereby they are initially sensitive to viewpoint and later encode view-invariant object identity. This transition depended on two aspects of object structure: it was strongest when objects foreshortened strongly across views and were similar to each other. View invariance in IT neurons was present even when objects were reduced to silhouettes, suggesting that it can arise through similarity between external contours of objects across views. Our results elucidate the viewpoint debate by showing that view invariance arises dynamically in IT neurons out of a representation that is initially view dependent.

First Structural View of a Peptide Interacting with the Nucleotide Binding Domain of Heat Shock Protein 90

The involvement of Hsp90 in progression of diseases like cancer, neurological disorders and several pathogen related conditions is well established. Hsp90, therefore, has emerged as an attractive drug target for many of these diseases. Several small molecule inhibitors of Hsp90, such as geldanamycin derivatives, that display antitumor activity, have been developed and are under clinical trials. However, none of these tested inhibitors or drugs are peptide-based compounds. Here we report the first crystal structure of a peptide bound at the ATP binding site of the N-terminal domain of Hsp90. The peptide makes several specific interactions with the binding site residues, which are comparable to those made by the nucleotide and geldanamycin. A modified peptide was designed based on these interactions. Inhibition of ATPase activity of Hsp90 was observed in the presence of the modified peptide. This study provides an alternative approach and a lead peptide molecule for the rational design of effective inhibitors of Hsp90 function.

Limited-view light sheet fluorescence microscopy for three dimensional volume imaging

We propose and demonstrate a limited-view light sheet microscopy (LV-LSM) for three dimensional (3D) volume imaging. Realizing that longer and frequent image acquisition results in significant photo-bleaching, we have taken limited angular views (18 views) of the macroscopic specimen and integrated with maximum likelihood (ML) technique for reconstructing high quality 3D volume images. Existing variants of light-sheet microscopy require both rotation and translation with a total of approximately 10-fold more views to render a 3D volume image. Comparatively, LV-LSM technique reduces data acquisition time and consequently minimizes light-exposure by many-folds. Since ML is a post-processing technique and highly parallelizable, this does not cost precious imaging time. Results show noise-free and high contrast volume images when compared to the state-of-the-art selective plane illumination microscopy. (C) 2015 AIP Publishing LLC.

ON THE UTILITY OF CANONICAL CORRELATION ANALYSIS FOR DOMAIN ADAPTATION IN MULTI-VIEW HEADPOSE ESTIMATION

The utility of canonical correlation analysis (CCA) for domain adaptation (DA) in the context of multi-view head pose estimation is examined in this work. We consider the three problems studied in 1], where different DA approaches are explored to transfer head pose-related knowledge from an extensively labeled source dataset to a sparsely labeled target set, whose attributes are vastly different from the source. CCA is found to benefit DA for all the three problems, and the use of a covariance profile-based diagonality score (DS) also improves classification performance with respect to a nearest neighbor (NN) classifier.

A dynamic view of the spread and intracellular distribution of Salmonella enterica.

The events that determine the dynamics of proliferation, spread and distribution of microbial pathogens within their hosts are surprisingly heterogeneous and poorly understood. We contend that understanding these phenomena at a sophisticated level with the help of mathematical models is a prerequisite for the development of truly novel, targeted preventative measures and drug regimes. We describe here recent studies of Salmonella enterica infections in mice which suggest that bacteria resist the antimicrobial environment inside host cells and spread to new sites, where infection foci develop, and thus avoid local escalation of the adaptive immune response. We further describe implications for our understanding of the pathogenic mechanism inside the host.

Field of view expansion for 3-D holographic display using a single spatial light modulator with scanning reconstruction light

Increasing the field of view of a holographic display while maintaining adequate image size is a difficult task. To address this problem, we designed a system that tessellates several sub-holograms into one large hologram at the output. The sub-holograms we generate is similar to a kinoform but without the paraxial approximation during computation. The sub-holograms are loaded onto a single spatial light modulator consecutively and relayed to the appropriate position at the output through a combination of optics and scanning reconstruction light. We will review the method of computer generated hologram and describe the working principles of our system. Results from our proof-of-concept system are shown to have an improved field of view and reconstructed image size. ©2009 IEEE.