We propose a holistic approach to the problem of re-identification in an environment of distributed smart cameras. We model the re-identification process in a distributed camera network as a distributed multi-class classifier, composed of spatially distributed binary classifiers. We treat the problem of re-identification as an open-world problem, and address novelty detection and forgetting. As there are many tradeoffs in design and operation of such a system, we propose a set of evaluation measures to be used in addition to the recognition performance. These are identity fragmentation, unknown rate and classifier coverage. The proposed concept is illustrated and evaluated on a new many-camera surveillance dataset and SAIVT-SoftBio dataset.
COBISS.SI-ID: 10896980
We analyzed a month-long measurement of indoor air and environment quality parameters in one gym during sporting activities such as football, basketball, volleyball, badminton, boxing and fitness. We have determined an average single person's contribution to the increase of temperature, humidity and dust concentration in the gym air volume of 12500 m3: during 90 min exercise performed at an average heart rate of 143 ± 10 bpm, a single person evaporated 0.94 kg of water into the air by sweating, contributed 0.03 K to the air temperature rise and added 1.5 μg/m3 and 5 ng/m3 to the indoor concentration of inhalable particles (PM10) and Ca concentration, respectively. As the air exchange by breathing at the observed exercise intensity was about three times faster with respect to the resting condition and since the exercise-induced PM10 concentration was about two times larger than outdoors, a sportsman in the gym would receive about a six-fold higher dose of PM10 inside than he/she would have received at rest outside.
COBISS.SI-ID: 1538041540
A novel, non-contact, optical type of surface moisture sensor was designed and investigated. It is based on absorption of NIR light absorption from laser source. Two laser sources are used, one with water-absorbing wavelength and the other which is much less affected by water and two InGaAs photo diodes. The sensor’s response was measured by a precision embedded data acquisition system NI CompactRIO and evaluated using a moisture analyser.
COBISS.SI-ID: 10861652
The paper describes development of the Slovenian emotional speech database for its primary use in speech synthesis. We also explore the potential of additional annotation for extending it for the use in emotion recognition tasks. The paper focus in methodology for annotating para lingual speaker information on the example of annotating speaker emotions in Slovenian radio dramas. Emotional speech database EmoLUKS was built from speech material of 17 Slovenian radio dramas. We obtained them from the national radio-and-television station (RTV Slovenia), which were given to the universities disposal with an academic license for processing and annotating the audio material. The utterances of one male and one female speaker were transcribed, segmented and then annotated with emotional states. The annotation of the emotional states was conducted in two stages with our own web-based application for crowd sourcing. Annotating assessments in different time periods with same five volunteers allows us to compare the obtained annotator’s decisions, therefore we report about annotator’s decisions consistency. Based on annotators majority vote of each annotated utterance we label speech material and join it to emotional speech database named Emo- LUKS. The material currently consists of 1385 recordings from one male (975 recordings) and one female (410 recordings) speaker and contains labeled emotional speech with a total duration of around 1 hour and 15 minutes. The paper presents the two-stage annotation process used to label the data and demonstrates the usefulness of used annotationmethodology. We evaluate the consistency of the annotated speechmaterial with the speaker dependent automatic emotion recognition system. The reported results are presented with the un-weighted as well as weighted average recalls and precisions for 2-class and 7-class recognition experiments. Results additionally confirms our presumption, that emotional speech database despite its complexity includes also clearly expressed emotional speaker states.
COBISS.SI-ID: 39227949
The invention proposes a new method and equipment for monitoring suitability of a tubular object geometry, which guides a user in properly positioning said complex-shaped object before photographing it and makes it possible to simply and most appropriately illuminate it and to rapidly assess the suitability of the object geometry. The invention is adapted for a fast industrial application. A computer provides for displaying where and in which position the tubular object to be monitored must be positioned, whereby higher accuracy of the proposed method is guaranteed. Identifying spots, on which the currently monitored tubular object critically deviates from the designed object, as well as automatic calibration of the equipment without need to position a calibration pattern next to the monitored object advantageously accelerate the monitoring procedure. An international patent application has been filed for this patent, which exhibits a significant potential for commercial exploitation.
COBISS.SI-ID: 11299156