Search results for: mapping algorithm

Authors: Siyamak Ziyaei, Siti Khalijah Mazlan, Petros Lappas

Abstract:

Compressed Natural Gas (CNG) mainly consists of Methane CH₄ and has a low carbon to hydrogen ratio relative to other hydrocarbons. As a result, it has the potential to reduce CO₂ emissions by more than 20% relative to conventional fuels like diesel or gasoline Although Natural Gas (NG) has environmental advantages compared to other hydrocarbon fuels whether they are gaseous or liquid, its main component, CH₄, burns at a slower rate than conventional fuels A higher pressure and a leaner cylinder environment will overemphasize slow burn characteristic of CH₄. Lean combustion and high compression ratios are well-known methods for increasing the efficiency of internal combustion engines. In order to achieve successful CNG lean combustion in Spark Ignition (SI) engines, a strong ignition system is essential to avoid engine misfires, especially in ultra-lean conditions. Turbulent Jet Ignition (TJI) is an ignition system that employs a pre-combustion chamber to ignite the lean fuel mixture in the main combustion chamber using a fraction of the total fuel per cycle. TJI enables ultra-lean combustion by providing distributed ignition sites through orifices. The fast burn rate provided by TJI enables the ordinary SI engine to be comparable to other combustion systems such as Homogeneous Charge Compression Ignition (HCCI) or Controlled Auto-Ignition (CAI) in terms of thermal efficiency, through the increased levels of dilution without the need of sophisticated control systems. Due to the physical geometry of TJIs, which contain small orifices that connect the prechamber to the main chamber, scavenging is one of the main factors that reduce TJI performance. Specifically, providing the right mixture of fuel and air has been identified as a key challenge. The reason for this is the insufficient amount of air that is pushed into the pre-chamber during each compression stroke. There is also the problem that combustion residual gases such as CO₂, CO and NOx from the previous combustion cycle dilute the pre- chamber fuel-air mixture preventing rapid combustion in the pre-chamber. An air-controlled active TJI is presented in this paper in order to address these issues. By applying air to the pre-chamber at a sufficient pressure, residual gases are exhausted, and the air-fuel ratio is controlled within the pre-chamber, thereby improving the quality of combustion. This paper investigates the 3D-simulated combustion characteristics of a Direct Injected (DI-CNG) fuelled SI en- gine with a pre-chamber equipped with an air channel by using AVL FIRE software. Experiments and simulations were performed at the Worldwide Mapping Point (WWMP) at 1500 Revolutions Per Minute (RPM), 3.3 bar Indicated Mean Effective Pressure (IMEP), using only conventional spark plugs as the baseline. After validating simulation data, baseline engine conditions were set for all simulation scenarios at λ=1. Following that, the pre-chambers with and without an auxiliary fuel supply were simulated. In the simulated (DI-CNG) SI engine, active TJI was observed to perform better than passive TJI and spark plug. In conclusion, the active pre-chamber with an air channel demon-strated an improved thermal efficiency (ηth) over other counterparts and conventional spark ignition systems.

Keywords: turbulent jet ignition, active air control turbulent jet ignition, pre-chamber ignition system, active and passive pre-chamber, thermal efficiency, methane combustion, internal combustion engine combustion emissions

Procedia PDF Downloads 87

Authors: Alessandro Preziosi, Antonio Prioletti, Luca Castangia

Abstract:

Deformable part models achieve high precision in pedestrian recognition, but all publicly available implementations are too slow for real-time applications. We implemented a deformable part model algorithm fast enough for real-time use by exploiting information about the camera position and orientation. This implementation is both faster and more precise than alternative DPM implementations. These results are obtained by computing convolutions in the frequency domain and using lookup tables to speed up feature computation. This approach is almost an order of magnitude faster than the reference DPM implementation, with no loss in precision. Knowing the position of the camera with respect to horizon it is also possible prune many hypotheses based on their size and location. The range of acceptable sizes and positions is set by looking at the statistical distribution of bounding boxes in labelled images. With this approach it is not needed to compute the entire feature pyramid: for example higher resolution features are only needed near the horizon. This results in an increase in mean average precision of 5% and an increase in speed by a factor of two. Furthermore, to reduce misdetections involving small pedestrians near the horizon, input images are supersampled near the horizon. Supersampling the image at 1.5 times the original scale, results in an increase in precision of about 4%. The implementation was tested against the public KITTI dataset, obtaining an 8% improvement in mean average precision over the best performing DPM-based method. By allowing for a small loss in precision computational time can be easily brought down to our target of 100ms per image, reaching a solution that is faster and still more precise than all publicly available DPM implementations.

Keywords: autonomous vehicles, deformable part model, dpm, pedestrian detection, real time

Procedia PDF Downloads 281

Authors: W. Hamaidia, T. Zebbiche

Abstract:

This paper presents the design of axisymmetric supersonic nozzles, in order to accelerate a supersonic flow to the desired Mach number and that having a small weight, in the same time gives a high thrust. The concerned nozzle gives a parallel and uniform flow at the exit section. The nozzle is divided into subsonic and supersonic regions. The supersonic portion is independent to the upstream conditions of the sonic line. The subsonic portion is used to give a sonic flow at the throat. In this case, nozzle gives a uniform and parallel flow at the exit section. It’s named by minimum length Nozzle. The study is done at high temperature, lower than the dissociation threshold of the molecules, in order to improve the aerodynamic performances. Our aim consists of improving the performances both by the increase of exit Mach number and the thrust coefficient and by reduction of the nozzle's mass. The variation of the specific heats with the temperature is considered. The design is made by the Method of Characteristics. The finite differences method with predictor-corrector algorithm is used to make the numerical resolution of the obtained nonlinear algebraic equations. The application is for air. All the obtained results depend on three parameters which are exit Mach number, the stagnation temperature, the chosen mesh in characteristics. A numerical simulation of nozzle through Computational Fluid Dynamics-FASTRAN was done to determine and to confirm the necessary design parameters.

Keywords: flux supersonic flow, axisymmetric minimum length nozzle, high temperature, method of characteristics, calorically imperfect gas, finite difference method, trust coefficient, mass of the nozzle, specific heat at constant pressure, air, error

Procedia PDF Downloads 150

Authors: J. S. Saini, P. P. K. Sandhu

Abstract:

The multi-hop nature of Wireless Mesh Networks and the hasty progression of throughput demands results in multi- channels and multi-radios structures in mesh networks, but the main problem of co-channels interference reduces the total throughput, specifically in multi-hop networks. Quality of Service mentions a vast collection of networking technologies and techniques that guarantee the ability of a network to make available desired services with predictable results. Quality of Service (QoS) can be directed at a network interface, towards a specific server or router's performance, or in specific applications. Due to interference among various transmissions, the QoS routing in multi-hop wireless networks is formidable task. In case of multi-channel wireless network, since two transmissions using the same channel may interfere with each other. This paper has considered the Destination Sequenced Distance Vector (DSDV) routing protocol to locate the secure and optimised path. The proposed technique also utilizes the Lempel–Ziv–Welch (LZW) based lossless data compression and intra cluster data aggregation to enhance the communication between the source and the destination. The use of clustering has the ability to aggregate the multiple packets and locates a single route using the clusters to improve the intra cluster data aggregation. The use of the LZW based lossless data compression has ability to reduce the data packet size and hence it will consume less energy, thus increasing the network QoS. The MATLAB tool has been used to evaluate the effectiveness of the projected technique. The comparative analysis has shown that the proposed technique outperforms over the existing techniques.

Keywords: WMNS, QOS, flooding, collision avoidance, LZW, congestion control

Procedia PDF Downloads 338

Authors: Jonghoon Chun, Insung Kim, Jonghyun Lim, Gun Ro

Abstract:

We believe that it is possible to provide a solution to reduce patient safety accidents by displaying correct medical records and prescription information through interactive patient identification. Our system is based on the use of smart bands worn by patients and these bands communicate with the hybrid gateways which understand both BLE and Wifi communication protocols. Through the convergence of low-power Bluetooth (BLE) and hybrid gateway technology, which is one of short-range wireless communication technologies, we implement ‘Intelligent Patient Identification and Location Tracking System’ to prevent medical malfunction frequently occurring in medical institutions. Based on big data and IOT technology using MongoDB, smart band (BLE, NFC function) and hybrid gateway, we develop a system to enable two-way communication between medical staff and hospitalized patients as well as to store locational information of the patients in minutes. Based on the precise information provided using big data systems, such as location tracking and movement of in-hospital patients wearing smart bands, our findings include the fact that a patient-specific location tracking algorithm can more efficiently operate HIS (Hospital Information System) and other related systems. Through the system, we can always correctly identify patients using identification tags. In addition, the system automatically determines whether the patient is a scheduled for medical service by the system in use at the medical institution, and displays the appropriateness of the medical treatment and the medical information (medical record and prescription information) on the screen and voice. This work was supported in part by the Korea Technology and Information Promotion Agency for SMEs (TIPA) grant funded by the Korean Small and Medium Business Administration (No. S2410390).

Keywords: BLE, hybrid gateway, patient identification, IoT, safety management, smart band

Procedia PDF Downloads 311

Authors: Stephen Tullis, M. David DiDonato, Hong Sung Park

Abstract:

Rowing boats (shells) are often steered by a small rudder operated by one of the backward-facing rowers; the attention required of that athlete then slightly decreases the power that that athlete can provide. Reducing the steering distraction would then increase the overall boat speed. Races are straight 2000 m courses with each boat in a 13.5 m wide lane marked by small (~15 cm) widely-spaced (~10 m) buoys, and the boat trajectory is affected by both cross-currents and winds. An optical buoy recognition and tracking system has been developed that provides the boat’s location and orientation with respect to the lane edges. This information is provided to the steering athlete as either: a simple overlay on a video display, or fed to a simplified autopilot system giving steering directions to the athlete or directly controlling the rudder. The system is then effectively a “lane-assist” device but with small, widely-spaced lane markers viewed from a very shallow angle due to constraints on camera height. The image is captured with a lightweight 1080p webcam, and most of the image analysis is done in OpenCV. The colour RGB-image is converted to a grayscale using the difference of the red and blue channels, which provides good contrast between the red/yellow buoys and the water, sky, land background and white reflections and noise. Buoy detection is done with thresholding within a tight mask applied to the image. Robust linear regression using Tukey’s biweight estimator of the previously detected buoy locations is used to develop the mask; this avoids the false detection of noise such as waves (reflections) and, in particular, buoys in other lanes. The robust regression also provides the current lane edges in the camera frame that are used to calculate the displacement of the boat from the lane centre (lane location), and its yaw angle. The interception of the detected lane edges provides a lane vanishing point, and yaw angle can be calculated simply based on the displacement of this vanishing point from the camera axis and the image plane distance. Lane location is simply based on the lateral displacement of the vanishing point from any horizontal cut through the lane edges. The boat lane position and yaw are currently fed what is essentially a stripped down marine auto-pilot system. Currently, only the lane location is used in a PID controller of a rudder actuator with integrator anti-windup to deal with saturation of the rudder angle. Low Kp and Kd values decrease unnecessarily fast return to lane centrelines and response to noise, and limiters can be used to avoid lane departure and disqualification. Yaw is not used as a control input, as cross-winds and currents can cause a straight course with considerable yaw or crab angle. Mapping of the controller with rudder angle “overall effectiveness” has not been finalized - very large rudder angles stall and have decreased turning moments, but at less extreme angles the increased rudder drag slows the boat and upsets boat balance. The full system has many features similar to automotive lane-assist systems, but with the added constraints of the lane markers, camera positioning, control response and noise increasing the challenge.

Keywords: auto-pilot, lane-assist, marine, optical, rowing

Procedia PDF Downloads 132

Authors: Richard Draper

Abstract:

A supercomputer can be constructed from identical building blocks which are small parallel processors connected by a network referred to as the local network. The routers have unused ports which are used to interconnect the building blocks. These connections are referred to as the global network. The address space has a global and a local component (g, l). The conventional way to connect the building blocks is to connect (g, l) to (g’,l). If there are K blocks, this requires K global ports in each router. If a block is of size M, the result is a machine with KM routers having diameter two. To increase the size of the machine to 2K blocks, each router connects to only half of the other blocks. The result is a larger machine but also one with greater diameter. This is a crude description of how the network of the CRAY XC® is designed. In this paper, a family of interconnection networks using routers with K global and M local ports is defined. Coordinates are (c,d, p) and the global connections are (c,d,p)↔(c’,p,d) which swaps p and d. The network is denoted D3(K,M) and is called a Swapped Dragonfly. D3(K,M) has KM2 routers and has diameter three, regardless of the size of K. To produce a network of size KM2 conventionally, diameter would be an increasing function of K. The family of Swapped Dragonflies has other desirable properties: 1) D3(K,M) scales linearly in K and quadratically in M. 2) If L < K, D3(K,M) contains many copies of D3(L,M). 3) If L < M, D3(K,M) contains many copies of D3(K,L). 4) D3(K,M) can perform an all-to-all exchange in KM2+KM time which is only slightly more than the time to do a one-to-all. This paper makes several contributions. It is the first time that a swap has been used to define a linearly scalable family of networks. Structural properties of this new family of networks are thoroughly examined. A synchronizing packet header is introduced. It specifies the path to be followed and it makes it possible to define highly parallel communication algorithm on the network. Among these is an all-to-all exchange in time KM2+KM. To demonstrate the effectiveness of the swap properties of the network of the CRAY XC® and D3(K,16) are compared.

Keywords: all-to-all exchange, CRAY XC®, Dragonfly, interconnection network, packet switching, swapped network, topology

Procedia PDF Downloads 121

Authors: Juni C. Kim, Anna R. Robuck, Douglas I. Walker

Abstract:

The human exposome, which includes chemical exposures over the lifetime and their effects, is now recognized as an important measure for understanding human health; however, the complexity of the data makes the identification of environmental chemicals challenging. The goal of our project was to establish a computational workflow for the improved identification of environmental pollutants containing chlorine or bromine. Using the “pattern. search” function available in the R package NonTarget, we wrote a multifunctional script that searches mass spectral clusters from untargeted gas-chromatography high-resolution mass spectrometry (GC-HRMS) for the presence of spectra consistent with chlorine and bromine-containing organic compounds. The “pattern. search” function was incorporated into a different function that allows the evaluation of clusters containing multiple analyte fragments, has multi-core support, and provides a simplified output identifying listing compounds containing chlorine and/or bromine. The new function was able to process 46,000 spectral clusters in under 8 seconds and identified over 150 potential halogenated spectra. We next applied our function to a deidentified dataset from patients diagnosed with primary biliary cholangitis (PBC), primary sclerosing cholangitis (PSC), and healthy controls. Twenty-two spectra corresponded to potential halogenated compounds in the PSC and PBC dataset, including six significantly different in PBC patients, while four differed in PSC patients. We have developed an improved algorithm for detecting halogenated compounds in GC-HRMS data, providing a strategy for prioritizing exposures in the study of human disease.

Keywords: exposome, metabolome, computational metabolomics, high-resolution mass spectrometry, exposure, pollutants

Procedia PDF Downloads 138

Authors: Yufeng Li, Yinan Ma, Jing Wu, Chengnian Long

Abstract:

Infrared object detection aims at separating small and dim target from clutter background and its capabilities extend beyond the limits of visible light, making it invaluable in a wide range of applications such as improving safety, security, efficiency, and functionality. However, existing methods are usually sensitive to the noise of the input infrared image, leading to a decrease in target detection accuracy and an increase in the false alarm rate in high-noise environments. To address this issue, an infrared small target detection algorithm called YOLO-IR is proposed in this paper to improve the robustness to high infrared noise. To address the problem that high noise significantly reduces the clarity and reliability of target features in infrared images, we design a soft-threshold coordinate attention mechanism to improve the model’s ability to extract target features and its robustness to noise. Since the noise may overwhelm the local details of the target, resulting in the loss of small target features during depth down-sampling, we propose a deep and shallow feature fusion neck to improve the detection accuracy. In addition, because the generalized Intersection over Union (IoU)-based loss functions may be sensitive to noise and lead to unstable training in high-noise environments, we introduce a Wasserstein-distance based loss function to improve the training of the model. The experimental results show that YOLO-IR achieves a 5.0% improvement in recall and a 6.6% improvement in F1-score over existing state-of-art model.

Keywords: infrared small target detection, high noise, robustness, soft-threshold coordinate attention, feature fusion

Procedia PDF Downloads 73

Authors: Ethan James

Abstract:

Over one billion people worldwide suffer from some level of vision loss or blindness as a result of progressive retinal diseases. Many patients, particularly in developing areas, are incorrectly diagnosed or undiagnosed whatsoever due to unconventional diagnostic tools and screening methods. Artificial intelligence (AI) based on deep learning (DL) convolutional neural networks (CNN) have recently gained a high interest in ophthalmology for its computer-imaging diagnosis, disease prognosis, and risk assessment. Optical coherence tomography (OCT) is a popular imaging technique used to capture high-resolution cross-sections of retinas. In ophthalmology, DL has been applied to fundus photographs, optical coherence tomography, and visual fields, achieving robust classification performance in the detection of various retinal diseases including macular degeneration, diabetic retinopathy, and retinitis pigmentosa. However, there is no complete diagnostic model to analyze these retinal images that provide a diagnostic accuracy above 90%. Thus, the purpose of this project was to develop an AI model that utilizes machine learning techniques to automatically diagnose specific retinal diseases from OCT scans. The algorithm consists of neural network architecture that was trained from a dataset of over 20,000 real-world OCT images to train the robust model to utilize residual neural networks with cyclic pooling. This DL model can ultimately aid ophthalmologists in diagnosing patients with these retinal diseases more quickly and more accurately, therefore facilitating earlier treatment, which results in improved post-treatment outcomes.

Keywords: artificial intelligence, deep learning, imaging, medical devices, ophthalmic devices, ophthalmology, retina

Procedia PDF Downloads 181

Authors: Erifyli Tsagkari, Stephanie Connelly, Zhaowei Liu, Andrew McBride, William Sloan

Abstract:

Biofilms play an essential role in treating water in biofiltration systems. The biofilm morphology and function are inextricably linked to the hydrodynamics of flow through a filter, and yet engineers rarely explicitly engineer this interaction. We develop a system that links computer simulation and 3-D printing to optimize and rapidly prototype filter media to optimize biofilm function with the hypothesis that biofilm function is intimately linked to the flow passing through the filter. A computational model that numerically solves the incompressible time-dependent Navier Stokes equations coupled to a model for biofilm growth and function is developed. The model is imbedded in an optimization algorithm that allows the model domain to adapt until criteria on biofilm functioning are met. This is applied to optimize the shape of filter media in a simple flow channel to promote biofilm formation. The computer code links directly to a 3-D printer, and this allows us to prototype the design rapidly. Its validity is tested in flow visualization experiments and by microscopy. As proof of concept, the code was constrained to explore a small range of potential filter media, where the medium acts as an obstacle in the flow that sheds a von Karman vortex street that was found to enhance the deposition of bacteria on surfaces downstream. The flow visualization and microscopy in the 3-D printed realization of the flow channel validated the predictions of the model and hence its potential as a design tool. Overall, it is shown that the combination of our computational model and the 3-D printing can be effectively used as a design tool to prototype filter media to optimize biofilm formation.

Keywords: biofilm, biofilter, computational model, von karman vortices, 3-D printing.

Procedia PDF Downloads 142

Authors: Lamees Nasser, Yago Diez, Robert Martí, Joan Martí, Ibrahim Sadek

Abstract:

Automated 3D breast ultrasound (ABUS) screening is a novel modality in medical imaging because of its common characteristics shared with other ultrasound modalities in addition to the three orthogonal planes (i.e., axial, sagittal, and coronal) that are useful in analysis of tumors. In the literature, few automatic approaches exist for typical tasks such as segmentation or registration. In this work, we deal with two problems concerning ABUS images: nipple and rib detection. Nipple and ribs are the most visible and salient features in ABUS images. Determining the nipple position plays a key role in some applications for example evaluation of registration results or lesion follow-up. We present a nipple detection algorithm based on color and shape of the nipple, besides an automatic approach to detect the ribs. In point of fact, rib detection is considered as one of the main stages in chest wall segmentation. This approach consists of four steps. First, images are normalized in order to minimize the intensity variability for a given set of regions within the same image or a set of images. Second, the normalized images are smoothed by using anisotropic diffusion filter. Next, the ribs are detected in each slice by analyzing the eigenvalues of the 3D Hessian matrix. Finally, a breast mask and a probability map of regions detected as ribs are used to remove false positives (FP). Qualitative and quantitative evaluation obtained from a total of 22 cases is performed. For all cases, the average and standard deviation of the root mean square error (RMSE) between manually annotated points placed on the rib surface and detected points on rib borders are 15.1188 mm and 14.7184 mm respectively.

Keywords: Automated 3D Breast Ultrasound, Eigenvalues of Hessian matrix, Nipple detection, Rib detection

Procedia PDF Downloads 330

Authors: Syed Jamil H. Kazmi, Faiza Sarwar

Abstract:

The Indus delta of Sindh Province forms the largest creeks zone of Pakistan. The Sindh coast starts from the mouth of Hab River and terminates at Sir Creek area. In this paper, we have considered the major creeks from the site of Bin Qasim Port in Karachi to Jetty of Keti Bunder in Thatta District. A general decline in the mangrove forest has been observed that within a span of last 25 years. The unprecedented human interventions damage the creeks habitat badly which includes haphazard urban development, industrial and sewage disposal, illegal cutting of mangroves forest, reduced and inconsistent fresh water flow mainly from Jhang and Indus rivers. These activities not only harm the creeks habitat but affected the fish stock substantially. Fishing is the main livelihood of coastal people but with the above-mentioned threats, it is also under enormous pressure by fish catches resulted in unchecked overutilization of the fish resources. This pressure is almost unbearable when it joins with deleterious fishing methods, uncontrolled fleet size, increase trash and by-catch of juvenile and illegal mesh size. Along with these anthropogenic interventions study area is under the red zone of tropical cyclones and active seismicity causing floods, sea intrusion, damage mangroves forests and devastation of fish stock. In order to sustain the natural resources of the Indus Creeks, this study was initiated with the support of FAO, WWF and NIO, the main purpose was to develop a Geo-Spatial dataset for fish stock assessment. The study has been spread over a year (2013-14) on monthly basis which mainly includes detailed fish stock survey, water analysis and few other environmental analyses. Environmental analysis also includes the habitat classification of study area which has done through remote sensing techniques for 22 years’ time series (1992-2014). Furthermore, out of 252 species collected, fifteen species from estuarine and marine groups were short-listed to measure the weight, health and growth of fish species at each creek under GIS data through SPSS system. Furthermore, habitat suitability analysis has been conducted by assessing the surface topographic and aspect derivation through different GIS techniques. The output variables then overlaid in GIS system to measure the creeks productivity. Which provided the results in terms of subsequent classes: extremely productive, highly productive, productive, moderately productive and less productive. This study has revealed the Geospatial tools utilization along with the evaluation of the fisheries resources and creeks habitat risk zone mapping. It has also been identified that the geo-spatial technologies are highly beneficial to identify the areas of high environmental risk in Sindh Creeks. This has been clearly discovered from this study that creeks with high rugosity are more productive than the creeks with low levels of rugosity. The study area has the immense potential to boost the economy of Pakistan in terms of fish export, if geo-spatial techniques are implemented instead of conventional techniques.

Keywords: fish stock, geo-spatial, productivity analysis, risk

Procedia PDF Downloads 245

Authors: Flavio Cannavo

Abstract:

Continuous evaluation of the status of potentially hazardous volcanos plays a key role for civil protection purposes. The importance of monitoring volcanic activity, especially for energetic paroxysms that usually come with tephra emissions, is crucial not only for exposures to the local population but also for airline traffic. Presently, real-time surveillance of most volcanoes worldwide is essentially delegated to one or more human experts in volcanology, who interpret data coming from different kind of monitoring networks. Unfavorably, the high nonlinearity of the complex and coupled volcanic dynamics leads to a large variety of different volcanic behaviors. Moreover, continuously measured parameters (e.g. seismic, deformation, infrasonic and geochemical signals) are often not able to fully explain the ongoing phenomenon, thus making the fast volcano state assessment a very puzzling task for the personnel on duty at the control rooms. With the aim of aiding the personnel on duty in volcano surveillance, here we introduce a system based on an ensemble of data-driven classifiers to infer automatically the ongoing volcano status from all the available different kind of measurements. The system consists of a heterogeneous set of independent classifiers, each one built with its own data and algorithm. Each classifier gives an output about the volcanic status. The ensemble technique allows weighting the single classifier output to combine all the classifications into a single status that maximizes the performance. We tested the model on the Mt. Etna (Italy) case study by considering a long record of multivariate data from 2011 to 2015 and cross-validated it. Results indicate that the proposed model is effective and of great power for decision-making purposes.

Keywords: Bayesian networks, expert system, mount Etna, volcano monitoring

Procedia PDF Downloads 246

Authors: S. S. Kharintsev, E. A. Chernykh, S. K. Saikin, A. I. Fishman, S. G. Kazarian

Abstract:

Recent advances in improving information storage performance are inseparably linked with circumvention of fundamental constraints such as the supermagnetic limit in heat assisted magnetic recording, charge loss tolerance in solid-state memory and the Abbe’s diffraction limit in optical storage. A substantial breakthrough in the development of nonvolatile storage devices with dimensional scaling has been achieved due to phase-change chalcogenide memory, which nowadays, meets the market needs to the greatest advantage. A further progress is aimed at the development of versatile nonvolatile high-speed memory combining potentials of random access memory and archive storage. The well-established properties of light at the nanoscale empower us to use them for recording optical information with ultrahigh density scaled down to a single molecule, which is the size of a pit. Indeed, diffraction-limited optics is able to record as much information as ~1 Gb/in2. Nonlinear optical effects, for example, two-photon fluorescence recording, allows one to decrease the extent of the pit even more, which results in the recording density up to ~100 Gb/in2. Going beyond the diffraction limit, due to the sub-wavelength confinement of light, pushes the pit size down to a single chromophore, which is, on average, of ~1 nm in length. Thus, the memory capacity can be increased up to the theoretical limit of 1 Pb/in2. Moreover, the field confinement provides faster recording and readout operations due to the enhanced light-matter interaction. This, in turn, leads to the miniaturization of optical devices and the decrease of energy supply down to ~1 μW/cm². Intrinsic features of light such as multimode, mixed polarization and angular momentum in addition to the underlying optical and holographic tools for writing/reading, enriches the storage and encryption of optical information. In particular, the finite extent of the near-field penetration, falling into a range of 50-100 nm, gives the possibility to perform 3D volume (layer-to-layer) recording/readout of optical information. In this study, we demonstrate a comprehensive evidence of isotropic-to-homeotropic phase transition of the azobenzene-functionalized polymer thin film exposed to light and dc electric field using near-field optical microscopy and scanning capacitance microscopy. We unravel a near-field Raman dichroism of a sub-10 nm thick epoxy-based side-chain azo-polymer films with polarization-controlled tip-enhanced Raman scattering. In our study, orientation of azo-chromophores is controlled with a bias voltage gold tip rather than light polarization. Isotropic in-plane and homeotropic out-of-plane arrangement of azo-chromophores in glassy environment can be distinguished with transverse and longitudinal optical near-fields. We demonstrate that both phases are unambiguously visualized by 2D mapping their local dielectric properties with scanning capacity microscopy. The stability of the polar homeotropic phase is strongly sensitive to the thickness of the thin film. We make an analysis of α-transition of the azo-polymer by detecting a temperature-dependent phase jump of an AFM cantilever when passing through the glass temperature. Overall, we anticipate further improvements in optical storage performance, which approaches to a single molecule level.

Keywords: optical memory, azo-dye, near-field, tip-enhanced Raman scattering

Procedia PDF Downloads 177

Authors: Keerti Kembhavi, M. R. Archana, V. Anjaneyappa

Abstract:

Most asphalt pavement condition evaluation use rating frameworks in which asphalt pavement distress is estimated by type, extent, and severity. Rating is carried out by the pavement condition rating (PCR), which is tedious and expensive. This paper presents the development of a low-cost technique for image pavement distress analysis that permits the identification of pothole and cracks. The paper explores the application of image processing tools for the detection of potholes and cracks. Longitudinal cracking and pothole are detected using Fuzzy-C- Means (FCM) and proceeded with the Spectral Theory algorithm. The framework comprises three phases, including image acquisition, processing, and extraction of features. A digital camera (Gopro) with the holder is used to capture pavement distress images on a moving vehicle. FCM classifier and Spectral Theory algorithms are used to compute features and classify the longitudinal cracking and pothole. The Matlab2016Ra Image preparing tool kit utilizes performance analysis to identify the viability of pavement distress on selected urban stretches of Bengaluru city, India. The outcomes of image evaluation with the utilization semi-computerized image handling framework represented the features of longitudinal crack and pothole with an accuracy of about 80%. Further, the detected images are validated with the actual dimensions, and it is seen that dimension variability is about 0.46. The linear regression model y=1.171x-0.155 is obtained using the existing and experimental / image processing area. The R2 correlation square obtained from the best fit line is 0.807, which is considered in the linear regression model to be ‘large positive linear association’.

Keywords: crack detection, pothole detection, spectral clustering, fuzzy-c-means

Procedia PDF Downloads 181

Authors: Bhat Geetalaxmi Jairam, D. V. Ashoka

Abstract:

Wireless Sensor Networks are used in many applications to collect sensed data from different sources. Sensed data has to be delivered through sensors wireless interface using multi-hop communication towards the sink. The data collection in wireless sensor networks consumes energy. Energy consumption is the major constraints in WSN .Reducing the energy consumption while increasing the amount of generated data is a great challenge. In this paper, we have implemented two data gathering protocols with multiple mobile sinks/elements to collect data from sensor nodes. First, is Energy-Efficient Data Gathering with Tour Length-Constrained Mobile Elements in Wireless Sensor Networks (EEDG), in which mobile sinks uses vehicle routing protocol to collect data. Second is An Intelligent Agent-based Routing Structure for Mobile Sinks in WSNs (IAR), in which mobile sinks uses prim’s algorithm to collect data. Authors have implemented concepts which are common to both protocols like deployment of mobile sinks, generating visiting schedule, collecting data from the cluster member. Authors have compared the performance of both protocols by taking statistics based on performance parameters like Delay, Packet Drop, Packet Delivery Ratio, Energy Available, Control Overhead. Authors have concluded this paper by proving EEDG is more efficient than IAR protocol but with few limitations which include unaddressed issues likes Redundancy removal, Idle listening, Mobile Sink’s pause/wait state at the node. In future work, we plan to concentrate more on these limitations to avail a new energy efficient protocol which will help in improving the life time of the WSN.

Keywords: aggregation, consumption, data gathering, efficiency

Procedia PDF Downloads 497

Authors: David C. Ni

Abstract:

Contemporary models for N-body systems are based on temporal, two-body, and mass point representation of Newtonian mechanics. Other mainstream models include 2D and 3D Ising models based on local neighborhood the lattice structures. In Quantum mechanics, the theories of collective modes are for superconductivity and for the long-range quantum entanglement. However, these models are still mainly for the specific phenomena with a set of designated parameters. We are therefore motivated to develop a new construction directly from the complex-variable N-body systems based on the extended Blaschke functions (EBF), which represent a non-temporal and nonlinear extension of Lorentz transformation on the complex plane – the normalized momentum spaces. A point on the complex plane represents a normalized state of particle momentums observed from a reference frame in the theory of special relativity. There are only two key parameters, normalized momentum and nonlinearity for modelling. An algorithm similar to Jenkins-Traub method is adopted for solving EBF iteratively. Through iteration, the solution sets show a form of σ + i [-t, t], where σ and t are the real numbers, and the [-t, t] shows various distributions, such as 1-peak, 2-peak, and 3-peak etc. distributions and some of them are analog to the canonical distributions. The results of the numerical analysis demonstrate continuum-to-discreteness transitions, evolutional invariance of distributions, phase transitions with conjugate symmetry, etc., which manifest the construction as a potential candidate for the unification of statistics. We hereby classify the observed distributions on the finite convergent domains. Continuous and discrete distributions both exist and are predictable for given partitions in different regions of parameter-pair. We further compare these distributions with canonical distributions and address the impacts on the existing applications.

Keywords: blaschke, lorentz transformation, complex variables, continuous, discrete, canonical, classification

Procedia PDF Downloads 309

Authors: Geetha Baskaran, Andrzej Barjiela, Rong Qu

Abstract:

Motivation: Nurse scheduling is a complex combinatorial optimization problem. It is also known as NP-hard. It needs an efficient re-scheduling to minimize some trade-off of the measures of violation by reducing selected constraints to soft constraints with measurements of their violations. Problem Statement: In this paper, we extend our novel approach to solve the nurse scheduling problem by transforming it through Information Granulation. Approach: This approach satisfies the rules of a typical hospital environment based on a standard benchmark problem. Generating good work schedules has a great influence on nurses' working conditions which are strongly related to the level of a quality health care. Domain transformation that combines the strengths of operation research and artificial intelligence was proposed for the solution of the problem. Compared to conventional methods, our approach involves judicious grouping (information granulation) of shifts types’ that transforms the original problem into a smaller solution domain. Later these schedules from the smaller problem domain are converted back into the original problem domain by taking into account the constraints that could not be represented in the smaller domain. An Integer Programming (IP) package is used to solve the transformed scheduling problem by expending the branch and bound algorithm. We have used the GNU Octave for Windows to solve this problem. Results: The scheduling problem has been solved in the proposed formalism resulting in a high quality schedule. Conclusion: Domain transformation represents departure from a conventional one-shift-at-a-time scheduling approach. It offers an advantage of efficient and easily understandable solutions as well as offering deterministic reproducibility of the results. We note, however, that it does not guarantee the global optimum.

Keywords: domain transformation, nurse scheduling, information granulation, artificial intelligence, simulation

Procedia PDF Downloads 397

Authors: Muhammad Arslan Usman, Muhammad Rehan Usman, Soo Young Shin

Abstract:

Live video streaming is one of the most widely used service among end users, yet it is a big challenge for the network operators in terms of quality. The only way to provide excellent Quality of Experience (QoE) to the end users is continuous monitoring of live video streaming. For this purpose, there are several objective algorithms available that monitor the quality of the video in a live stream. Subjective tests play a very important role in fine tuning the results of objective algorithms. As human perception is considered to be the most reliable source for assessing the quality of a video stream, subjective tests are conducted in order to develop more reliable objective algorithms. Temporal impairments in a live video stream can have a negative impact on the end users. In this paper we have conducted subjective evaluation tests on a set of video sequences containing temporal impairment known as frame freezing. Frame Freezing is considered as a transmission error as well as a hardware error which can result in loss of video frames on the reception side of a transmission system. In our subjective tests, we have performed tests on videos that contain a single freezing event and also for videos that contain multiple freezing events. We have recorded our subjective test results for all the videos in order to give a comparison on the available No Reference (NR) objective algorithms. Finally, we have shown the performance of no reference algorithms used for objective evaluation of videos and suggested the algorithm that works better. The outcome of this study shows the importance of QoE and its effect on human perception. The results for the subjective evaluation can serve the purpose for validating objective algorithms.

Keywords: objective evaluation, subjective evaluation, quality of experience (QoE), video quality assessment (VQA)

Procedia PDF Downloads 601

Authors: Marina Benito-Parejo, Raul Perez-Lopez, Miguel Herraiz, Carolina Guardiola-Albert, Cesar Martinez

Abstract:

Persistency, long-term memory and randomness are intrinsic properties of time-series of earthquakes. The Rescaled Range Analysis (RS-Analysis) was introduced by Hurst in 1956 and modified by Mandelbrot and Wallis in 1964. This method represents a simple and elegant analysis which determines the range of variation of one natural property (the seismic energy released in this case) in a time interval. Despite the simplicity, there is complexity inherent in the property measured. The cumulative curve of the energy released in time is the well-known fractal geometry of a devil’s staircase. This geometry is used for determining the maximum and minimum value of the range, which is normalized by the standard deviation. The rescaled range obtained obeys a power-law with the time, and the exponent is the Hurst value. Depending on this value, time-series can be classified in long-term or short-term memory. Hence, an algorithm has been developed for compiling the RS-Analysis for time series of earthquakes by days. Completeness time distribution and locally stationarity of the time series are required. The interest of this analysis is their application for a complex seismic crisis where different earthquakes take place in clusters in a short period. Therefore, the Hurst exponent has been obtained for the seismic crisis of Alhoceima (Mediterranean Sea) of January-March, 2016, where at least five medium-sized earthquakes were triggered. According to the values obtained from the Hurst exponent for each cluster, a different mechanical origin can be detected, corroborated by the focal mechanisms calculated by the official institutions. Therefore, this type of analysis not only allows an approach to a greater understanding of a seismic series but also makes possible to discern different types of seismic origins.

Keywords: Alhoceima crisis, earthquake time series, Hurst exponent, rescaled range analysis

Procedia PDF Downloads 321

Authors: Ho Yeon Park, Kyoung-Jae Kim

Abstract:

Recently, user’s interests for location-based social network service increases according to the advances of social web and location-based technologies. It may be easy to recommend preferred items if we can use user’s preference, context and social network information simultaneously. In this study, we propose context-aware POI (point-of-interests) recommender systems using location-based network analysis and sentiment analysis which consider context, social network information and implicit user’s preference score. We propose a context-aware POI recommendation system consisting of three sub-modules and an integrated recommendation system of them. First, we will develop a recommendation module based on network analysis. This module combines social network analysis and cluster-indexing collaboration filtering. Next, this study develops a recommendation module using social singular value decomposition (SVD) and implicit SVD. In this research, we will develop a recommendation module that can recommend preference scores based on the frequency of POI visits of user in POI recommendation process by using social and implicit SVD which can reflect implicit feedback in collaborative filtering. We also develop a recommendation module using them that can estimate preference scores based on the recommendation. Finally, this study will propose a recommendation module using opinion mining and emotional analysis using data such as reviews of POIs extracted from location-based social networks. Finally, we will develop an integration algorithm that combines the results of the three recommendation modules proposed in this research. Experimental results show the usefulness of the proposed model in relation to the recommended performance.

Keywords: sentiment analysis, network analysis, recommender systems, point-of-interests, business analytics

Procedia PDF Downloads 250

Authors: Azadeh Jafari, Robert G. Owens

Abstract:

In this study, a geometrical multiscale approach which means coupling together the 2-D Navier-Stokes equations, constitutive equations and 0-D lumped parameter models is investigated. A multiscale approach, suggest a natural way of coupling detailed local models (in the flow domain) with coarser models able to describe the dynamics over a large part or even the whole cardiovascular system at acceptable computational cost. In this study we introduce a new velocity correction scheme to decouple the velocity computation from the pressure one. To evaluate the capability of our new scheme, a comparison between the results obtained with Neumann outflow boundary conditions on the velocity and Dirichlet outflow boundary conditions on the pressure and those obtained using coupling with the lumped parameter model has been performed. Comprehensive studies have been done based on the sensitivity of numerical scheme to the initial conditions, elasticity and number of spectral modes. Improvement of the computational algorithm with stable convergence has been demonstrated for at least moderate Weissenberg number. We comment on mathematical properties of the reduced model, its limitations in yielding realistic and accurate numerical simulations, and its contribution to a better understanding of microvascular blood flow. We discuss the sophistication and reliability of multiscale models for computing correct boundary conditions at the outflow boundaries of a section of the cardiovascular system of interest. In this respect the geometrical multiscale approach can be regarded as a new method for solving a class of biofluids problems, whose application goes significantly beyond the one addressed in this work.

Keywords: geometrical multiscale models, haemorheology model, coupled 2-D navier-stokes 0-D lumped parameter modeling, computational fluid dynamics

Procedia PDF Downloads 361

Authors: Mahesh K. Jat, Ankan Jana, Mahender Choudhary

Abstract:

Rapid population and economic growth resulted in changes in large-scale land use land cover (LULC) changes. Changes in the biophysical properties of the Earth's surface and its impact on climate are of primary concern nowadays. Different approaches, ranging from location-based relationships or modelling earth surface - atmospheric interaction through modelling techniques like surface energy balance (SEB) are used in the recent past to examine the relationship between changes in Earth surface land cover and climatic characteristics like temperature and precipitation. A remote sensing-based model i.e., Surface Energy Balance Algorithm for Land (SEBAL), has been used to estimate the surface heat fluxes over Mahi Bajaj Sagar catchment (India) from 2001 to 2020. Landsat ETM and OLI satellite data are used to model the SEB of the area. Changes in observed precipitation and temperature, obtained from India Meteorological Department (IMD) have been correlated with changes in surface heat fluxes to understand the relative contributions of LULC change in changing these climatic variables. Results indicate a noticeable impact of LULC changes on climatic variables, which are aligned with respective changes in SEB components. Results suggest that precipitation increases at a rate of 20 mm/year. The maximum and minimum temperature decreases and increases at 0.007 ℃ /year and 0.02 ℃ /year, respectively. The average temperature increases at 0.009 ℃ /year. Changes in latent heat flux and sensible heat flux positively correlate with precipitation and temperature, respectively. Variation in surface heat fluxes influences the climate parameters and is an adequate reason for climate change. So, SEB modelling is helpful to understand the LULC change and its impact on climate.

Keywords: LULC, sensible heat flux, latent heat flux, SEBAL, landsat, precipitation, temperature

Procedia PDF Downloads 116

Authors: Maryam Zarabian, Hector Guzman, Pedro Pereira-Almao, Abraham Fapojuwo

Abstract:

Dry reforming of methane (DRM) has sparked significant industrial and scientific interest not only as a viable alternative for addressing the environmental concerns of two main contributors of the greenhouse effect, i.e., carbon dioxide (CO₂) and methane (CH₄), but also produces syngas, i.e., a mixture of hydrogen (H₂) and carbon monoxide (CO) utilized by a wide range of downstream processes as a feedstock for other chemical productions. In this study, we develop an AI-enable syngas production model to tackle the problem of achieving an equivalent H₂/CO ratio [1:1] with respect to the most efficient conversion. Firstly, the unsupervised density-based spatial clustering of applications with noise (DBSAN) algorithm removes outlier data points from the original experimental dataset. Then, random forest (RF) and deep neural network (DNN) models employ the error-free dataset to predict the DRM results. DNN models inherently would not be able to obtain accurate predictions without a huge dataset. To cope with this limitation, we employ reusing pre-trained layers’ approaches such as transfer learning and greedy layer-wise pretraining. Compared to the other deep models (i.e., pure deep model and transferred deep model), the greedy layer-wise pre-trained deep model provides the most accurate prediction as well as similar accuracy to the RF model with R² values 1.00, 0.999, 0.999, 0.999, 0.999, and 0.999 for the total outlet flow, H₂/CO ratio, H₂ yield, CO yield, CH₄ conversion, and CO₂ conversion outputs, respectively.

Keywords: artificial intelligence, dry reforming of methane, artificial neural network, deep learning, machine learning, transfer learning, greedy layer-wise pretraining

Procedia PDF Downloads 86

Authors: An Chengrui, Yin Zi, Wu Bingbing, Ma Yuanzhu, Jin Kaixiu, Chen Xiao, Ouyang Hongwei

Abstract:

Single cell RNA sequence (scRNA-seq) is one of the effective tools to study transcriptomics of biological processes. Recently, similarity measurement of cells is Euclidian distance or its derivatives. However, the process of scRNA-seq is a multi-variate Bernoulli event model, thus we hypothesize that it would be more efficient when the divergence between cells is valued with relative entropy than Euclidian distance. In this study, we compared the performances of Euclidian distance, Spearman correlation distance and Relative Entropy using scRNA-seq data of the early, medial and late stage of limb development generated in our lab. Relative Entropy is better than other methods according to cluster potential test. Furthermore, we developed KL-SNE, an algorithm modifying t-SNE whose definition of divergence between cells Euclidian distance to Kullback–Leibler divergence. Results showed that KL-SNE was more effective to dissect cell heterogeneity than t-SNE, indicating the better performance of relative entropy than Euclidian distance. Specifically, the chondrocyte expressing Comp was clustered together with KL-SNE but not with t-SNE. Surprisingly, cells in early stage were surrounded by cells in medial stage in the processing of KL-SNE while medial cells neighbored to late stage with the process of t-SNE. This results parallel to Heatmap which showed cells in medial stage were more heterogenic than cells in other stages. In addition, we also found that results of KL-SNE tend to follow Gaussian distribution compared with those of the t-SNE, which could also be verified with the analysis of scRNA-seq data from another study on human embryo development. Therefore, it is also an effective way to convert non-Gaussian distribution to Gaussian distribution and facilitate the subsequent statistic possesses. Thus, relative entropy is potentially a better way to determine the divergence of cells in scRNA-seq data analysis.

Keywords: Single cell RNA sequence, Similarity measurement, Relative Entropy, KL-SNE, t-SNE

Procedia PDF Downloads 340

Authors: Maher Ali Rusho, Sudipta Halder

Abstract:

The purpose of the study is to develop optimal solutions for models of parallel computing systems using the Java language. During the study, programmes were written for the examined models of parallel computing systems. The result of the parallel sorting code is the output of a sorted array of random numbers. When processing data in parallel, the time spent on processing and the first elements of the list of squared numbers are displayed. When processing requests asynchronously, processing completion messages are displayed for each task with a slight delay. The main results include the development of optimisation methods for algorithms and processes, such as the division of tasks into subtasks, the use of non-blocking algorithms, effective memory management, and load balancing, as well as the construction of diagrams and comparison of these methods by characteristics, including descriptions, implementation examples, and advantages. In addition, various specialised libraries were analysed to improve the performance and scalability of the models. The results of the work performed showed a substantial improvement in response time, bandwidth, and resource efficiency in parallel computing systems. Scalability and load analysis assessments were conducted, demonstrating how the system responds to an increase in data volume or the number of threads. Profiling tools were used to analyse performance in detail and identify bottlenecks in models, which improved the architecture and implementation of parallel computing systems. The obtained results emphasise the importance of choosing the right methods and tools for optimising parallel computing systems, which can substantially improve their performance and efficiency.

Keywords: algorithm optimisation, memory management, load balancing, performance profiling, asynchronous programming.

Procedia PDF Downloads 12

Authors: Dario Barsi, Marina Ubaldi, Pietro Zunino, Robert Fink

Abstract:

A design and optimization procedure is proposed and developed to provide the geometry of a high efficiency compact hydraulic propeller turbine for low head. For the preliminary design of the machine, classic design criteria, based on the use of statistical correlations for the definition of the fundamental geometric parameters and the blade shapes are used. These relationships are based on the fundamental design parameters (i.e., specific speed, flow coefficient, work coefficient) in order to provide a simple yet reliable procedure. Particular attention is paid, since from the initial steps, on the correct conformation of the meridional channel and on the correct arrangement of the blade rows. The preliminary geometry thus obtained is used as a starting point for the hydrodynamic optimization procedure, carried out using a CFD calculation software coupled with a genetic algorithm that generates and updates a large database of turbine geometries. The optimization process is performed using a commercial approach that solves the turbulent Navier Stokes equations (RANS) by exploiting the axial-symmetric geometry of the machine. The geometries generated within the database are therefore calculated in order to determine the corresponding overall performance. In order to speed up the optimization calculation, an artificial neural network (ANN) based on the use of an objective function is employed. The procedure was applied for the specific case of a propeller turbine with an innovative design of a modular type, specific for applications characterized by very low heads. The procedure is tested in order to verify its validity and the ability to automatically obtain the targeted net head and the maximum for the total to total internal efficiency.

Keywords: renewable energy conversion, hydraulic turbines, low head hydraulic energy, optimization design

Procedia PDF Downloads 150

Authors: Abolghasem Yousefi-Babadi, Ali Bozorgi-Amiri, Aida Kazempour, Reza Tavakkoli-Moghaddam, Maryam Irani

Abstract:

All over the world, natural disasters (e.g., earthquakes, floods, volcanoes and hurricanes) causes a lot of deaths. Earthquakes are introduced as catastrophic events, which is accident by unusual phenomena leading to much loss around the world. Such could be replaced by disasters or any other synonyms strongly demand great long-term help and relief, which can be hard to be managed. Supplies and facilities are very important challenges after any earthquake which should be prepared for the disaster regions to satisfy the people's demands who are suffering from earthquake. This paper proposed disaster response facility allocation problem for disaster relief operations as a mathematical programming model. Not only damaged people in the earthquake victims, need the consumable commodities (e.g., food and water), but also they need non-consumable commodities (e.g., clothes) to protect themselves. Therefore, it is concluded that paying attention to disaster points and people's demands are very necessary. To deal with this objective, both commodities including consumable and need non-consumable commodities are considered in the presented model. This paper presented the multi-objective multi-period mathematical programming model regarding the minimizing the average of the weighted response times and minimizing the total operational cost and penalty costs of unmet demand and unused commodities simultaneously. Furthermore, a Chebycheff multi-objective solution procedure as a powerful solution algorithm is applied to solve the proposed model. Finally, to illustrate the model applicability, a case study of the Tehran earthquake is studied, also to show model validation a sensitivity analysis is carried out.

Keywords: facility location, multi-objective model, disaster response, commodity

Procedia PDF Downloads 257

Authors: Anthony D. Rhodes, Manan Goel

Abstract:

We provide a high fidelity deep learning algorithm (HyperSeg) for interactive video segmentation tasks using a dense convolutional network with context-aware skip connections and compressed, 'hypercolumn' image features combined with a convolutional tessellation procedure. In order to maintain high output fidelity, our model crucially processes and renders all image features in high resolution, without utilizing downsampling or pooling procedures. We maintain this consistent, high grade fidelity efficiently in our model chiefly through two means: (1) we use a statistically-principled, tensor decomposition procedure to modulate the number of hypercolumn features and (2) we render these features in their native resolution using a convolutional tessellation technique. For improved pixel-level segmentation results, we introduce a boundary loss function; for improved temporal coherence in video data, we include temporal image information in our model. Through experiments, we demonstrate the improved accuracy of our model against baseline models for interactive segmentation tasks using high resolution video data. We also introduce a benchmark video segmentation dataset, the VFX Segmentation Dataset, which contains over 27,046 high resolution video frames, including green screen and various composited scenes with corresponding, hand-crafted, pixel-level segmentations. Our work presents a improves state of the art segmentation fidelity with high resolution data and can be used across a broad range of application domains, including VFX pipelines and medical imaging disciplines.

Keywords: computer vision, object segmentation, interactive segmentation, model compression

Procedia PDF Downloads 120