GPU based building footprint identification utilising self-attention multiresolution analysis

Purpose: Accurate and fast delineation of targets and normal structures based on daily CTs is critical for online adaptive radiotherapy (ART). Manual contouring generally takes too long to be practical. Auto-segmentation can run in a few minutes on modern computers, but it is desirable to reduce this time further. A simple, inexpensive option to accelerate auto-segmentation is to use a graphics processor unit (GPU). This work aims to evaluate the performance of a GPU-accelerated auto-segmentation. Method and Materials: A GPU card (NVidia GTX 285) was added to a computer consisting of two 3GHz four core CPUs running auto-segmentation software, ABAS (Atlas Based Auto Segmentation, Elekta CMS Software). To evaluate performance, this setup was tested using daily CTs acquired during IGRT using a CT-on-Rails (CTVision, Siemens) for 14 breast (4 prone, 10 supine) and 6 prostate cancer patients. Contours of targets and critical structures were populated from planning CTs to daily CTs using ABAS with and without the GPU. The accuracy of GPU-accelerated contouring was measured using Dice's Coefficient (DC). Results: For prostate cases, there is no noticeable difference between contours generated with and without GPU (<2% difference in DC to physician contours). The average time to segment was reduced by 30%, to 1.4 minutes, with GPU, (as low as 0.97 minutes). For prone breast cases, except for lumpectomy cavity (DC: 77 – 93%), there is also no substantial difference between contours with or without GPU (DC>95%). The average time to segment images was reduced by 33% with GPU (to 2.8 minutes). For supine breast cases, contours generated with and without GPU agreed equally with contours drawn by physicians. The GPU average segmentation time was 2.3 minutes, reduced by 47%. Conclusion: The use of GPU can substantially accelerate auto-segmentation without degrading the segmentation accuracy. This time saving is significant for online ART.

Modified local ternary patterns technique for brain tumour segmentation and volume estimation from MRI multi-sequence scans with GPU CUDA machine

In some applications, evolutionary algorithms may require high computational resources and high processing power, sometimes not producing a satisfactory solution after running for a considerable amount of time. One possible improvement is a parallel approach to reduce the response time. This work proposes to study a parallel multi-objective algorithm, the multi-objective version of Differential Evolution (DE). The generation of trial individuals can be done in parallel, greatly reducing the overall processing time of the algorithm. A novel approach to parallelize this algorithm is the implementation on the Graphic Processing Units (GPU). These units present high degree of parallelism and they were initially developed for image rendering. However, NVIDIA has released a framework, named CUDA, which allows developers to use GPU for general-purpose computing (GPGPU). This work studies the implementation of Multi-Objective DE (MODE) on the GPU with C-CUDA, evaluating the gain in processing time against the sequential version. Benchmark functions are used to validate the implementation and to confirm the efficiency of MODE on the GPU. The results show that the approach achieves an expressive speed up and a highly efficient processing power.

Person re-identification (Re-ID) aims to retrieve a person of interest across multiple nonoverlapping cameras. In the last few years, the construction of aerial person Re-ID datasets has been appealing because visual surveillance through unmanned aerial vehicle (UAV) platforms has become very valuable in real-world scenarios. However, great differences are exhibited between the pedestrian images captured by ground cameras and those captured by UAVs. Person Re-ID methods based on ground person images have difficulty when performing Re-ID based on aerial person images. In this paper, we first propose a novel meta-transfer learning method for person Re-ID in aerial imagery; this approach trains a generalisable Re-ID model to learn discriminative feature representations for aerial person images. Specifically, a meta-learning strategy is introduced to study a feature extractor, and a transfer learning strategy is introduced to utilise and further improve the acquired meta-knowledge. To overcome the convergence speed and recognition accuracy reductions caused by the presence of difficult categories in the given dataset, we propose a learning strategy based on curriculum sampling that is harmonised with our meta-transfer learning framework. In addition, a new metric formulation of sample similarity is introduced based on Mahalanobis distance to improve the optimisation of the model. Extensive comparative evaluation experiments are conducted on the large-scale aerial Re-ID dataset, and the results obtained show that our method achieves a Rank-1 accuracy of 63.63 $$\%$$ and a mean average precision (mAP) of 38.02 $$\%$$ , demonstrating its potential for completing person Re-ID in aerial images. Ablation studies also validate that each component contributes to improving the performance of the model.

Aerial imagery captured via unmanned aerial vehicles (UAVs) is playing an increasingly important role in disaster response. Unlike satellite imagery, aerial imagery can be captured and processed within hours rather than days. In addition, the spatial resolution of aerial imagery is an order of magnitude higher than the imagery produced by the most sophisticated commercial satellites today. Both the United States Federal Emergency Management Agency (FEMA) and the European Commission's Joint Research Center (JRC) have noted that aerial imagery will inevitably present a big data challenge. The purpose of this article is to get ahead of this future challenge by proposing a hybrid crowdsourcing and real-time machine learning solution to rapidly process large volumes of aerial data for disaster response in a time-sensitive manner. Crowdsourcing can be used to annotate features of interest in aerial images (such as damaged shelters and roads blocked by debris). These human-annotated features can then be used to train a supervised machine learning system to learn to recognize such features in new unseen images. In this article, we describe how this hybrid solution for image analysis can be implemented as a module (i.e., Aerial Clicker) to extend an existing platform called Artificial Intelligence for Disaster Response (AIDR), which has already been deployed to classify microblog messages during disasters using its Text Clicker module and in response to Cyclone Pam, a category 5 cyclone that devastated Vanuatu in March 2015. The hybrid solution we present can be applied to both aerial and satellite imagery and has applications beyond disaster response such as wildlife protection, human rights, and archeological exploration. As a proof of concept, we recently piloted this solution using very high-resolution aerial photographs of a wildlife reserve in Namibia to support rangers with their wildlife conservation efforts (SAVMAP project, http://lasig.epfl.ch/savmap ). The results suggest that the platform we have developed to combine crowdsourcing and machine learning to make sense of large volumes of aerial images can be used for disaster response.

Large scale parallelisation of the material point method with multiple GPUs

Rooftop edge line extraction is essential for building recognition and extraction in aerial images. In this paper, a new integrated edge line extraction algorithm is proposed for aerial imagery which often contains many buildings of complex shapes, and much more noise. Firstly, an improved marker-controlled watershed algorithm is applied to the aerial imagery segmentation, incorporated with altitude information of Digital Surface Model (DSM) to remove roads and shades. Meanwhile, the Canny operator is used to detect the edges of buildings. Then, for each building region obtained by segmentation, the new algorithm applies the line detection method based on phase-grouping to edge pixels. After edge lines are extracted initially, rooftop constraints are introduced in the rooftop main orientation counting procedure, which could help design and adjust lines. With all the procedures above, experiments show that the proposed method could achieve efficient and effective extraction of building edge line feature in aerial images.

The Single Instruction Multiple Data (SIMD) architecture of Graphic Processing Units (GPUs) makes them perfect for parallel processing of big data. In this paper, we present the design, implementation and evaluation of G-Storm , a GPU-enabled parallel system based on Storm, which harnesses the massively parallel computing power of GPUs for high-throughput online stream data processing. G-Storm has the following desirable features: 1) G-Storm is designed to be a general data processing platform as Storm, which can handle various applications and data types. 2) G-Storm exposes GPUs to Storm applications while preserving its easy-to-use programming model. 3) G-Storm achieves high-throughput and low-overhead data processing with GPUs. 4) G-Storm accelerates data processing further by enabling Direct Data Transfer (DDT), between two executors that process data at a common GPU. We implemented G-Storm based on Storm 0.9.2 and tested it using three different applications, including continuous query, matrix multiplication and image resizing. Extensive experimental results show that 1) Compared to Storm, G-Storm achieves over 7× improvement on throughput for continuous query, while maintaining reasonable average tuple processing time. It also leads to 2.3× and 1.3× throughput improvements on the other two applications, respectively. 2) DDT significantly reduces data processing time.

This contribution shows how unsupervised Markovian segmentation techniques can be accelerated when implemented on graphics hardware equipped with a Graphics Processing Unit (GPU). Our strategy exploits the intrinsic properties of local interactions between sites of a Markov Random Field model with the parallel computation ability of a GPU. This paper explains how classical iterative site-wise-update algorithms commonly used to optimize global Markovian cost functions can be efficiently implemented in parallel by fragment shaders driven by a fragment processor. This parallel programming strategy significantly accelerates optimization algorithms such as ICM and simulated annealing. Good acceleration are also achieved for parameter estimation procedures such as K-means and ICE. The experiments reported in this paper have been obtained with a mid-end, affordable graphics card available on the market.

Recently, IoT (Internet of Things) technologies have been progressed. To overcome of the high cost of developing IoT services by vertically integrating devices and services, Open IoT enables various IoT services to be developed by integrating horizontally separated devices and services. For Open IoT, we have proposed Tacit Computing technology to discover the devices that have data users need on demand and use them dynamically and an automatic GPU (graphics processing unit) offloading technology as an elementary technology of Tacit Computing. However, it can improve limited applications because it only optimizes parallelizable loop statements extraction. Therefore, in this paper, to improve performances of more applications automatically, we propose an improved method with reduction of data transfer between CPU and GPU. This can improve performance of many IoT applications. We evaluate our proposed GPU offloading method by applying it to Darknet which is general large application for CPU and find that it can process it 3 times as quickly as only using CPUs within 10 h tuning time.

Summary form only given. Present day computers equipped with powerful graphics processing units (GPUs) show considerable promise of increased performance for the electromagnetic (EM) modeler. In order to determine the degree of performance gain achievable for electro-energetic physics computations, the MAGIC EM finite difference-time domain (FDTD) particle-in-cell (PIC) plasma code is undergoing testing for parallel speedup on typical large scale plasma wave EM calculations using GPU processing. The results of the tests are expected to be of great interest to the EM community because MAGIC is widely employed and implements the methods common to many large scale applications. The performance data obtained will quantify the benefits of GPUs, and aid in selecting the proper system including GPUs potentially critical to effectiveness with EM/plasma codes where calculation times often run into days due to small zone and time step requirements. GPUs promise a relatively low cost boost leveraging the vast development driven by the computer game industry.

DOAJ is a unique and extensive index of diverse open access journals from around the world, driven by a growing community, committed to ensuring quality content is freely available online for everyone.

4D MR phase and magnitude segmentations with GPU parallel computing

Application of graphics processing unit (GPU) to software in elementary particle/high energy physics field

Aim. Determination of the elements of external spatial orientation of the surveying systems at the moment of image acquisition is the fundamental task in photogrammetry. Principally, this problem is solving in two ways. The first way is direct positioning and measuring of directions of camera optical axis in the geodetic space with the help of GNSS/INS equipment. The second way is the analytical solution of the problem using a set of reference information (often such information is a set of ground control points whose geodetic positions are known with sufficient accuracy and which are reliably recognised on aerial images of the photogrammetric block). The authors consider the task of providing reference and control information using the second approach, which has a number of advantages in terms of reliability and accuracy of determining the unknown image exterior orientation parameters. It is proposed to obtain additional images of ground control points by the method of their auxiliary aerial photography using an unmanned aerial vehicle (UAV) on a larger scale compared to the scale of the images of the photogrammetric block. The aim of the presented work is the implementation of the method of creating reference points and experimental confirmation of its effectiveness for photogrammetric processing. Methods and results. For the entire realization of the potential of the analytical way to determine the elements of external orientation of images, it is necessary to have a certain number of ground control points (GCP) and to keep the defined scheme of their location on the photogrammetric block. As the main source of input data authors use UAV aerial images of the terrain, which are obtained separately from the block of aerial survey, and have a better geometric resolution and which clearly depict the control reference points. Application of such auxiliary images gives the possibility of automated transferring of the position of ground control point into images of the main photogrammetric block. In our interpretation, these images of ground control points and their surroundings on the ground are called "control reference images". The basis of the work is to develop a method for obtaining the auxiliary control reference images and transferring of position of GCP depicted on them into aerial or space images of terrain by means of computer stereo matching. To achieve this goal, we have developed a processing method for the creation of control reference images of aerial image or a series of auxiliary multi-scale aerial images obtained by a drone from different heights above the reference point. The operator identifies and measures the GCP once on the auxiliary aerial image of the highest resolution. Then there is an automatic stereo matching of the control reference image in the whole series of auxiliary images in succession with a decrease in the resolution and, ultimately, directly with the aerial images of photogrammetric block. On this stage there are no recognition/cursor targeting by the human operator, and therefore there are no discrepancies, errors or mistakes related to it. In addition, if to apply fairly large size of control reference images, the proposed method can be used on a low-texture terrain, and therefore deal in many cases without the physical marking of points measured by GNSS method. And this is a way to simplify and reduce the cost of photogrammetric technology. The action of the developed method has been verified experimentally to provide the control reference information of the block of archival aerial images of the low-texture terrain. The results of the experimental approbation of the proposed method give grounds to assert that the method makes it possible to perform geodetic reference of photogrammetric projects more efficiently due to the refusal of the physical marking of the area before aerial survey. The proposed method can also be used to obtain the information for checking the quality of photogrammetric survey for provision of check points. The authors argue that the use of additional equipment - UAV of semi-professional class to obtain control reference images is economically feasible. Scientific novelty and practical relevance. The results of approbation of the "control reference image" method with obtaining stereo pairs of aerial images with vertical placement of the base are presented for the first time. There was implemented the study of the properties of such stereo pairs of aerial images to obtain images of reference points. The effectiveness of including reference images in the main block of the digital aerial triangulation network created on UAV’s images is shown.