With the emergence of Convolutional Neural Network (CNN) models, precision of image classification tasks has been improved significantly over these years. Regional CNN (RCNN) model is proposed to solve object detection tasks with a combination of Region Proposal Network and CNN. This model improves the detection accuracy but suffer from slow inference speed because of its multi-stage structure. The Single Stage Detection (SSD) network is later proposed to further improve the object detection benchmark in terms of accuracy and speed. However, SSD model still suffers from high miss rate on small targets since datasets are usually dominated by medium and large sized objects, which don’t share the same features with small ones. On the other hand, geometric analysis on dataset images can provide additional information before model training. In this thesis, we propose several SSD-based models with adjusted parameters on feature extraction layers by using geometric analysis on KITTI and Caltech Pedestrian datasets. This analysis extends SSD’s capability on small objects detection. To further improve detection accuracy, we propose a two-stream network, which uses one stream to detect medium to large objects, and another stream specifically for small objects. This two-stream model achieves competitive performance comparing to other algorithms on KITTI and Caltech Pedestrian benchmark. Those results are shown and analysed in this thesis as well.

In this thesis, we improved our understanding of the effects of hydrology, stream hydraulics, and temperature on juvenile Atlantic salmon (Salmo salar L.) habitat. We demonstrated the key role played by stream morphology and flow regime on in-channel hydraulics and consequently on salmonid habitat. Additionally, we showed how riparian afforestation has potential to moderate climate change effects on stream temperature preserving freshwater ecosystems. The Girnock Burn is an upland Scottish river that has been intensively studied to investigate flow generation processes and stream temperature, and has served as a long-term monitoring site for Atlantic salmon population dynamics since 1966. The general approach applied consisted of combining different types of models including: hydraulic, fish habitat, hydrological and heat transfer models with long-term hydrological and climatic data sets, and digital terrain models (DTMs) at different spatio-temporal scales. Our results showed that the extensive presence of roughness elements (e.g. boulders and cobbles) is capable to provide some refuge areas for juvenile salmon fry for a wide range of flows. However, under extreme flow conditions, in-channel hydraulics might represent a limiting factor. Significant inter-site differences occurred and were consistent throughout the years. Evidence of long-term trend in fry habitat quality could be identified only in summer. Since more extreme flow regimes are expected in the future as a result of climate change, we also proposed a novel analytical approach that allowed us to assess the effects of hydroclimatic variation on fish populations outside the range of observations. Finally, we showed the potential of afforestation to reduce daily stream temperature range, moderating both low and high peaks of more than 2 ○C. This makes riparian shading a valuable mitigation strategy to contrast global warming effects on stream temperatures that should be considered for a sustainable catchment management.

Práce shrnuje problematiku digitálního videa na WEBu a nastiňuje možnosti využití videa, jako velice účinného způsobu přenosu informací po internetu. Dále práce identifikuje způsoby využití digitálního videa v praxi a pro jednotlivé způsoby použití stanovuje doporučení týkající se konkrétních charakteristik digitálního videa. Vyjadřuje se také k budoucím trendům v oblasti digitálního videa na WEBu. Pro porozumění závěrečných kapitol jsou nejdříve rozebrány obecné principy digitálního videa. Proto se v první části práce zaměřuje na digitální video obecně a s ním spojené pojmy. Poté jsou rozebírány způsoby získávání záznamů pro internetové použití. Následuje souhrn možných využití videa na webu. V další části jsou rozebrány nejpoužívanější formáty a s nimi spojené používané algoritmy komprese. Celá jedna kapitola je věnovaná problematice řízení digitálních práv ? DRM. Poté se podíváme na platformy pro běžné přehrávání po internetu využívající streaming videa, jakožto velice perspektivní technologie. Následně jsou zmíněny některé aspekty distribuce digitálního videa a je analyzován stav internetových videopůjčoven v ČR. Práce dále identifikuje požadavky jednotlivých způsobů použití a je navrženo řešení se zohledněním celé problematiky, tzn. jaký formát a jaké nastavení na které platformě použít. V závěru je zhodnocena celá problematika a shrnuty stěžejní body práce. Nakonec jsou identifikovány trendy v oblasti digitálního videa a je předpovězen vývoj dané problematiky.

Class of 2006 Abstract / Objectives: To develop and evaluate a data collection form that could be used to assess the influence of heparin administration on the septic severity index score in patients with Candida blood stream infections. Methods: A data collection form was developed to evaluate the influence of heparin administration on the septic severity index score in patients with Candida blood stream infections. This form was assessed using rating scales for ease of use, applicability of data collection items and availability of data in charts. Results: Data from 10 patient charts was used to assess the instrument. Patient demographics were similar. The strengths of the form included applicability of 9 out of 10 items in the instrument. Four out of 10 items were rated low for ease of use. In addition, 7 out of 10 pertinent data items were not documented in the charts. Conclusion: An extensive reworking of the data collection form as described in the discussion section was required. The form is ready to be implemented if the proposed study should be performed.

The voluminous limnological literature contains few studies of mountain streams. Though there are a few papers on the benthic algae (see Budde 1928, Raabe 1951) only Pennak (1943) reports year-round quantitative data on the phytoplankton. Brinley (1950) gives some phytoplankton information in a summer study of streams in Rocky Mountain National Park.

In this study, options for restoring fish populations in Hamilton City (37.47'S, 175.19'E) were explored. Habitat and fish populations in Hamilton urban streams were manipulated using a two-fold experimental design. Firstly, habitat was enhanced in ten urban streams with three continuous treatments in a 60-m reach at each site (20 m with 10 ponga logs, 20 m with 20 hollow clay pipes, and 20 m with no added structure). Secondly, juvenile farm-reared giant kokopu (Galaxias argenteus), were stocked into five of the enhanced stream sections. Giant kokopu are threatened and occur naturally in Hamilton urban streams in sparse populations. The abundance of wild fish was monitored before and after enhancement and fish release from November 2006 to November 2007. Stocked fish were monitored for eight months, from April to November 2007. Over this time electric fishing was conducted three times, trap nets (Gee minnow and fyke nets) were set monthly and spotlighting was conducted monthly at three release sites where water clarity allowed. Anticipated outcomes of this research were; to determine whether giant kokopu abundance in Hamilton urban streams is limited by recruitment or by habitat, and to assist with the development of methods to restore fish populations in Hamilton City urban streams. Logs used as enhancement structures in Hamilton urban streams provided more stable habitat for fish and created more suitable microhabitat than pipe structures. Pipes moved considerably during high flows, and their instability made them less effective at providing habitat. Within the study sites there appeared to be complex interactions with turbidity, stream width and depth, which complicated the effect of the habitat structures. The limited replication and variability among sites contributed to statistically insignificant results using analysis of variance. Retention and recapture rates of stocked juvenile giant kokopu were greatest at Site M11, where the stream was narrow, shallow, clear and had lower numbers and biomass of shortfin eels, compared to other survey sites. Marked and released giant kokopu were retained in the release reaches at four of the five sites, for a minimum of four months, and exhibited substantial growth. Daily growth of juvenile giant kokopu ranged from 0.19 to 0.33 mm day-1 and from 0.03 to 0.11 g day-1, exhibiting substantial growth over winter. Giant kokopu appeared to have a slight bias to the log section of enhanced habitat, but habitat selection appeared to be overwhelmingly controlled by initial habitat selection. The stocking of farm-reared fish into urban streams was largely successful, but the success of the habitat enhancement was variable and further work is required to determine better techniques for habitat enhancement in these urban environments. It is concluded that releasing farm-reared giant kokopu can be used to restore populations especially where recruitment limitations control fish abundance and diversity.

n/a

The huge volume of time series data generated in many applications poses new challenges in the techniques of data storage, transmission, and computation. Further more, when the time series are in the form of streaming data, new problems emerge and new techniques are required because of the streaming characteristics, e.g. high volume, high speed and continuous flowing. Approximate representation is one of the most efficient and effective solutions to address the large-volume-high-speed problem. In this thesis, we propose a dynamic representation scheme for streaming time series. Existing methods use a unitary function form for the entire approximation task. In contrast, our method adopts a set of function candidates such as linear function, polynomial function(degree ≥ 2), and exponential function. We provide a novel segmenting strategy to generate subsequences and dynamically choose candidate functions to approximate the subsequences. / Since we are dealing with streaming time series, the segmenting points and the corresponding approximate functions are incrementally produced. For a certain function form, we use a buffer window to find the local farthest possible segmenting point under a user specified error tolerance threshold. To achieve this goal, we define a feasible space for the coefficients of the function and show that we can indirectly find the local best segmenting point by the calculation in the coefficient space. Given the error tolerance threshold, the candidate function representing more information by unit parameter is chosen as the approximate function. Therefore, our representation scheme is more flexible and compact. We provide two dynamic algorithms, PLQS and PLQES, which involve two and three candidate functions, respectively. We also present the general strategy of function selection when more candidate functions are considered. In the experimental test, we examine the effectiveness of our algorithms with synthetic and real time series data sets. We compare our method with the piecewise linear approximation method and the experimental results demonstrate the evident superiority of our dynamic approach under the same error tolerance threshold.

Thesis (M.S.)--University of Nevada, Reno, 2005. / "December, 2005." Includes bibliographical references (leaves 103-112). Online version available on the World Wide Web.

<p>Portable stream programming language (PSPL) is a language for baseband application</p><p>programming on reconfigurable architectures. The first step in its development has been</p><p>completed. A parser has been provided for the front end of the PSPL compiler. The syntax</p><p>of the language has been fixed to allow for easy parses. The scanner and the parser</p><p>where generated using automatic tools (scanner and parser generators) which rely on</p><p>complex mathematical algorithms for their generation. Abstract syntax (data structures</p><p>that preserve the source program so that program structure is evident) was implemented</p><p>for the parser using a syntax separate from interpretation style of programming. Tests were</p><p>carried out to ensure that the correct data structures were generated. The final outcome</p><p>is a parser that other phases of the compiler can depend on for onward transmission of</p><p>the source program in an unambiguous manner. The development of subsequent phases</p><p>of the compiler will form the next logical step in the processes of transforming PSPL to</p><p>a stand alone language.</p>