Evaluation of Localized Semantics: Data, Methodology, and Experiments

We present a new data set of 1014 images with manual segmentations and semantic labels for each segment, together with a methodology for using this kind of data for recognition evaluation. The images and segmentations are from the UCB segmentation benchmark database (Martin et al., in International conference on computer vision, vol. II, pp. 416–421, 2001). The database is extended by manually labeling each segment with its most specific semantic concept in WordNet (Miller et al., in Int. J. Lexicogr. 3(4):235–244, 1990). The evaluation methodology establishes protocols for mapping algorithm specific localization (e.g., segmentations) to our data, handling synonyms, scoring matches at different levels of specificity, dealing with vocabularies with sense ambiguity (the usual case), and handling ground truth regions with multiple labels. Given these protocols, we develop two evaluation approaches. The first measures the range of semantics that an algorithm can recognize, and the second measures the frequency that an algorithm recognizes semantics correctly. The data, the image labeling tool, and programs implementing our evaluation strategy are all available on-line (kobus.ca//research/data/IJCV_2007). We apply this infrastructure to evaluate four algorithms which learn to label image regions from weakly labeled data. The algorithms tested include two variants of multiple instance learning (MIL), and two generative multi-modal mixture models. These experiments are on a significantly larger scale than previously reported, especially in the case of MIL methods. More specifically, we used training data sets up to 37,000 images and training vocabularies of up to 650 words. We found that one of the mixture models performed best on image annotation and the frequency correct measure, and that variants of MIL gave the best semantic range performance. We were able to substantively improve the performance of MIL methods on the other tasks (image annotation and frequency correct region labeling) by providing an appropriate prior.     

Metal hydride hydrogen compressors: A review

Metal hydride (MH) thermal sorption compression is an efficient and reliable method allowing a conversion of energy from heat into a compressed hydrogen gas. The most important component of such a thermal engine – the metal hydride material itself – should possess several material features in order to achieve an efficient performance in the hydrogen compression. Apart from the hydrogen storage characteristics important for every solid H storage material (e.g. gravimetric and volumetric efficiency of H storage, hydrogen sorption kinetics and effective thermal conductivity), the thermodynamics of the metal–hydrogen systems is of primary importance resulting in a temperature dependence of the absorption/desorption pressures). Several specific features should be optimised to govern the performance of the MH-compressors including synchronisation of the pressure plateaus for multi-stage compressors, reduction of slope of the isotherms and hysteresis, increase of cycling stability and life time, together with challenges in system design associated with volume expansion of the metal matrix during the hydrogenation.     

A numerical Hartree-Fock program for diatomic molecules

This paper describes an algorithm and a computer program which solves numerically (virtually exactly) equations of the restricted open-shell Hartree-Fock and Hartree-Fock-Slater model for diatomic molecules     

An experimental investigation into forced, natural and combined forced and natural convective heat transfer from stationary isothermal circular disks

Experimental heat transfer data are presented and dimensionless correlations developed for forced, natural and combined assisting forced and natural convection for heated stationary isothermal circular disks over wide ranges of the Reynolds, Rayleigh and modified Reynolds numbers, respectively. Experiments with air were performed for a variety of disks ranging in diameter and thickness-to-diameter aspect ratio. The correlation for combined forced and natural convection was developed utilizing the concept of a modified Reynolds number which accounts for a buoyancy-induced velocity. Utilizing this concept, the experimental data and respective empirical correlations for all three convection modes can be collapsed and plotted on the same continuous curve.     

Phenolic compounds and antioxidant activity of extracts of Ginkgo leaves

The aim of the present study was to determine the flavonol contents of crude extracts from Ginkgo (Ginkgo biloba L.) leaves, their antioxidant properties in model system studies, and their inhibitory action of linoleic acid oxidation in an emulsion system and of triacylglycerols (TAG) in rapeseed oil. Extracts were prepared from both green and yellow leaves of Ginkgo trees using water, aqueous acetone, and ethanol. Extracts were analyzed by HPLC for their presence and content of selected flavonols; these included myricetin, quercetin, kaempferol, isorhamnetin, and morin. The highest level of flavonols, especially quercetin, myricetin and kaempferol, were found in the aqueous acetonic extracts. Crude extracts from yellow Ginkgo leaves contained greater amounts of flavonols. The best DPPH radical‐scavenging activity amongst the Ginkgo extracts examined was determined for aqueous acetonic extracts, while the lowest was noted for the ethanolic extract of green leaves. Water infusion extracts exhibited the highest iron(II) chelating activity. The reducing power of extracts from yellow leaves was 2 higher than that of crude extracts from green leaves. Nevertheless, extracts from green Ginkgo leaves imparted a greater protection factor against TAG oxidation, as assessed by the Rancimat method. Crude extracts from yellow Ginkgo leaves were more efficacious than green ones at inhibiting oxidation of the linoleic acid emulsion.     

Selection of metal hydrides-based thermal energy storage: Energy storage efficiency and density targets

Thermo-chemical energy storage based on metal hydrides has gained tremendous interest in solar heat storage applications such as concentrated solar power systems (CSP) and parabolic troughs. In such systems, two metal hydride beds are connected and operating in an alternative way as energy storage or hydrogen storage. However, the selection of metal hydrides is essential for a smooth operation of these CSP systems in terms of energy storage efficiency and density. In this study, thermal energy storage systems using metal hydrides are modeled and analyzed in detail using first law of thermodynamics. For these purpose, four conventional metal hydrides are selected namely LaNi₅, Mg, Mg₂Ni and Mg₂FeH₆. The comparison of performance is made in terms of volumetric energy storage and energy storage efficiency. The effects of operating conditions (temperature, hydrogen pressure and heat transfer fluid mass flow rates) and reactor design on the aforementioned performance metrics are studied and discussed in detail. The preliminary results showed that Mg-based hydrides store energy ranging from 1.3 to 2.4 GJ m^?3 while the energy storage can be as low as 30% due to their slow intrinsic kinetics. On the other hand, coupling Mg-based hydrides with LaNi₅ allow us to recover heat at a useful temperature above 330 K with low energy density ca.500 MJ m^?3 provided suitable operating conditions are selected. The results of this study will be helpful to screen out all potentially viable hydrides materials for heat storage applications.     

Thermal Decomposition of Partially Hydrogenated Rapeseed Oil During Repeated Frying Traditional and Fast French Fries

The aim of this study was to compare thermal degradation of oil, especially the composition of the polymer in a polar and nonpolar fraction of oil, used for repeated frying of fast and traditional French fries. The French fries were fried using the partially hydrogenated rapeseed oil. Fast French fries were characterized by a half shorter frying time compared to traditional ones. The frying process was done at 170 °C ± 5 °C in 5‐l electric fryers and carried out in 15‐min cycles for 48 hours. The content of thermal decomposition of triacylglycerol (TAG) in both fractions of oil was analyzed by high‐performance size‐exclusion chromatography (HPSEC). In all analyzed samples, thermal decomposition products were found. However, the composition of a polar and nonpolar fraction of oil was not the same. In a nonpolar fraction, only the monomers and hydrolysis products of TAG were observed. In a polar fraction, dimers, trimers, and oligomers of TAG were also found. The shorter time of frying the fast French fries resulted in a lower total and individual polymers content in all steps of analysis compared to the oil used for frying the traditional French fries.     

Near-field visualization of strongly confined surface plasmon polaritons in metal-insulator-metal waveguides

A nanoscale gap between two metal surfaces can confine propagating surface plasmon polaritons (SPPs) to very small dimensions, but this geometry makes it inherently difficult to image SPP propagation at high resolution. We demonstrate the near-field probing of these SPPs, propagating within a 50 nm thick Si 3N 4 waveguide with Ag cladding layers for frequencies ranging from the blue to the near-infrared. Using near-field SPP interferometry, we determine the wave vector, showing that the wavelength is shortened to values as small as 156 nm for a free-space wavelength of 532 nm.