Search

The security of women is of prime concern around the world. Women feel insecure while traveling out of the home due to the fear of violence. The fear of violence restricts women's participating in different social activities. So instead of becoming a victim of a violent crime such as domestic violence, robbery, or rape, women should call on resources to help her out of that situation. In this paper, the authors develop a women safety device, namely SafeWomen, which helps in reducing the crimes held against women. This is a new approach for providing security to women in any unsafe situation by sending an alert having geographical location along with emergency message to the registered contact numbers so that the incident could be prevented. Also, it can track the current location of the victim just by knowing the IP address of the device she is using. One can also use this system for the safety and security of kids and elderly people just by making some changes in the functionality of the system.

As it has become increasingly common for social network users to write and view post in languages other than English, most social networks now provide machine translations to allow posts to be read by an audience beyond native speakers. However, authors typically cannot view the translations of their posts and have little control over these translations. To address this issue, I am developing a prototype that will provide authors with transparency of and more personalized control over the translation of their posts.

We consider the problem of utility maximization in online ranking applications while also satisfying a pre-defined fairness constraint. We consider batches of items which arrive over time, already ranked using an existing ranking model. We propose online post-processing for re-ranking these batches to enforce adherence to the pre-defined fairness constraint, while maximizing a specific notion of utility. To achieve this goal, we propose two deterministic re-ranking policies. In addition, we learn a re-ranking policy based on a novel variation of learning to search. Extensive experiments on real world and synthetic datasets demonstrate the effectiveness of our proposed policies both in terms of adherence to the fairness constraint and utility maximization. Furthermore, our analysis shows that the performance of the proposed policies depends on the original data distribution w.r.t the fairness constraint and the notion of utility.

LiDAR provides highly accurate 3-D point clouds. However, data need to be manually labeled in order to provide subsequent useful information. Manual annotation of such data is time-consuming, tedious, and error prone, and hence, in this article, we present three automatic methods for annotating trees in LiDAR data. The first method requires high-density point clouds and uses certain LiDAR data attributes for the purpose of tree identification, achieving almost 90% accuracy. The second method uses a voxel-based 3-D convolutional neural network on low-density LiDAR data sets and is able to identify most large trees accurately but struggles with smaller ones due to the voxelization process. The third method is a scaled version of the PointNet++ method and works directly on outdoor point clouds and achieves an $F_{\mathrm{ score}}$ of 82.1% on the ISPRS benchmark data set, comparable to the state-of-the-art methods but with increased efficiency.

Satellite images are an extremely valuable resource in the aftermath of natural disasters such as hurricanes and tsunamis where they can be used for risk assessment and disaster management. In order to provide timely and actionable information for disaster response, in this paper a framework utilising segmentation neural networks is proposed to identify impacted areas and accessible roads in post-disaster scenarios. The effectiveness of pretraining with ImageNet on the task of aerial image segmentation has been analysed and performances of popular segmentation models compared. Experimental results show that pretraining on ImageNet usually improves the segmentation performance for a number of models. Open data available from OpenStreetMap (OSM) is used for training, forgoing the need for time-consuming manual annotation. The method also makes use of graph theory to update road network data available from OSM and to detect the changes caused by a natural disaster. Extensive experiments on data from the 2018 tsunami that struck Palu, Indonesia show the effectiveness of the proposed framework. ENetSeparable, with 30% fewer parameters compared to ENet, achieved comparable segmentation results to that of the state-of-the-art networks., Comment: Accepted in Neurocomputing, 2020

Explainability is an important factor to drive user trust in the use of neural networks for tasks with material impact. However, most of the work done in this area focuses on image analysis and does not take into account 3D data. We extend the saliency methods that have been shown to work on image data to deal with 3D data. We analyse the features in point clouds and voxel spaces and show that edges and corners in 3D data are deemed as important features while planar surfaces are deemed less important. The approach is model-agnostic and can provide useful information about learnt features. Driven by the insight that 3D data is inherently sparse, we visualise the features learnt by a voxel-based classification network and show that these features are also sparse and can be pruned relatively easily, leading to more efficient neural networks. Our results show that the Voxception-ResNet model can be pruned down to 5\% of its parameters with negligible loss in accuracy., Comment: Accepted for IJCNN 2020

Current methods for building tissues usually start with a non-biological blueprint, or rely on self-organization, which does not extend to organ-scales. This has limited the construction of large tissues that simultaneously encode fine-scale cell organization. Here we bridge scales by mimicking developmental dynamics using “kinomorphs”, tissue scaffolds that undergo globally programmed shape and density changes to trigger local self-organization of cells in many locations at once. In this first report, we focus on mimicking the extracellular matrix (ECM) compaction and division into leaflets that occurs in kidney collecting duct development. We start by creating single-cell resolution cell patterns in ECM-mimetic hydrogels that are >10x larger than previously described, by leveraging photo-lithographic technology. These patterns are designed to mimic the branch geometry of the embryonic kidney collecting duct tree. We then predict the shape dynamics of kinomorphs driven by cell contractility-based compaction of the ECM using kinematic origami simulations. We show that these dynamics spur centimeter-scale assembly of structurally mature ~50 μm-diameter epithelial tubules that are locally self-organized, but globally programmed. Our approach prescribes tubule network geometry at ~5x smaller length-scales than currently possible using 3D printing, and at local cell densities comparable to in vivo tissues. Kinomorphs could be used to scaffold and “plumb” arrays of organoids in the future, by guiding the morphogenesis of epithelial networks. Such hybrid globally programmed/locally self-organized tissues address a major gap in our ability to recapitulate organ-scale tissue structure.Significance StatementEngineers are attempting to build tissues that mimic human diseases outside of the body. Although stem cells can be coaxed to form small organoids with a diversity of cell types, they do not properly organize over large distances by themselves. We report a strategy to mimic developmental processes using dynamic materials that attempt to guide a cellular “blueprint” towards a more complex tissue endpoint. We call these materials kinomorphs, combining the Greek kinó (propel, drive) and morfí (form, shape), since they seek to shepherd both the shape and developmental trajectory of cell collectives within them. Kinomorphs could pave the way towards organ-scale synthetic tissues built through a hybrid of engineering and self-organization strategies.

Accurate road maps are important for timely disaster relief efforts and risk management. Current disaster mapping is done manually by volunteers following a disaster and the process is slow and error prone. We propose a framework for identifying accessible roads in post-disaster satellite imagery by detecting changes from pre-disaster imagery, in conjunction with OpenStreetMap data. We validate our results with data from Indonesia 2018 tsunami, obtained from DigitalGlobe.

Timely disaster risk management requires accurate road maps and prompt damage assessment. Currently, this is done by volunteers manually marking satellite imagery of affected areas but this process is slow and often error-prone. Segmentation algorithms can be applied to satellite images to detect road networks. However, existing methods are unsuitable for disaster-struck areas as they make assumptions about the road network topology which may no longer be valid in these scenarios. Herein, we propose a CNN-based framework for identifying accessible roads in post-disaster imagery by detecting changes from pre-disaster imagery. Graph theory is combined with the CNN output for detecting semantic changes in road networks with OpenStreetMap data. Our results are validated with data of a tsunami-affected region in Palu, Indonesia acquired from DigitalGlobe.

The detection of buried minimum-metal anti-personnel landmines is a time-consuming problem, due to the high false alarm rate (FAR) arising from metallic clutter typically found in minefields. Magnetic induction spectroscopy (MIS) offers a potential way to reduce the FAR by classifying the metallic objects into threat and non-threat categories, based on their spectroscopic signatures. A new algorithm for threat identification for MIS sensors, based on a fully-connected artificial neural network (ANN), is proposed in this paper, and compared against a classifier based on Support Vector Machines (SVM). The results demonstrate that MIS is a potentially viable option for the reduction of false alarms in humanitarian demining. It is also shown that the ANN outperforms the SVM-based approach for threat objects containing minimal amounts of metal.

In medical field, ultrasound imaging technique is a major medium for the doctors and radiologists to diagnose the patient’s problem. The presence of speckle noise degrades the visuals of the ultrasound image and making it difficult for the doctors to diagnose properly. Hence, denoising and restoration of these images came up as a challenge to enhance the quality of the image. In this paper, an improved non-local mean filter is proposed for suppression of speckle noise in ultrasound image. This improved filter is the combination of two conventional filters, i.e., non-local mean filter and bilateral filter. The newly modified method which is emerged by combining the functions of both the conventional filters results in the improvement of quality of image. The three image quality assessment parameters are being used to analyze the performance of improved non-local mean filter, i.e., peak signal-to-noise-ratio (PSNR) and structural similarity (SSIM).

A semantic feature extraction method for multitemporal high resolution aerial image registration is proposed in this paper. These features encode properties or information about temporally invariant objects such as roads and help deal with issues such as changing foliage in image registration, which classical handcrafted features are unable to address. These features are extracted from a semantic segmentation network and have shown good robustness and accuracy in registering aerial images across years and seasons in the experiments.

The suppression of speckle noise is necessary for clear vision of ultrasound images. The quality of ultrasound images is degraded by the presence of speckle noise. In this work, bilateral Filter is used to suppress speckle noise. Conventionally, this filter is used to suppress the Gaussian noise from the images. A bilateral filter is better at edge preserving, noise suppression and for better smoothening of gray as well as colored images. Bilateral filter tends to improve image quality as it replaces the intensity of each pixel with a weighted average of intensity values from nearby pixels. These weights are basically based on Gaussian distribution function. The three parameters have been used to analyze the performance of bilateral filter the are PSNR, SSIM, and SSI.

The speckle suppression is very important for carrying out proper clinical operation in ultrasound images. Speckle is a kind of multiplicative noise which exists inherently within the ultrasound images. Although there are many speckle noise suppression filters from the ultrasound images, it possesses certain constraints. In this paper, the Non-local Mean (NLM) filter is used to suppress speckle noise from ultrasound images. NLM filter consists of two windows that are search and similarity windows which can improve upon the limitation of many conventional filters. The simulation results of NLM filter for the values of different noise variances on ultrasound image are shown. Finally, the performance analysis is done by image quality assessment parameters like Peak Signal-to-Noise Ratio (PSNR) and Coefficient of Correlation (CoC).

LiDAR provides highly accurate 3D point cloud data for a number of tasks such as forest surveying and urban planning. Automatic classification of this data, however, is challenging since the dataset can be extremely large and manual annotation is labour intensive if not impossible. We provide a method of automatically annotating airborne LiDAR data for individual trees or tree regions by filtering out the ground measurements and then using the number of returns embedded in the dataset. The method is validated on a manually annotated dataset for Dublin city with promising results.

With the support of cellular system a cellular-based mobile ad hoc network (MANET) offers promising communication scenarios while entails secure data exchange as other wireless systems. In this paper, we propose a novel decentralized key generation mechanism using shared symmetric polynomials in which the base stations (BSs) carry out an initial key generation by a symmetric polynomial in a distributed manner and then pass on the key material to mobile stations (MSs). Thereafter, our proposed key generation scheme enables each pair of MSs to establish a pairwise key without any intervention from the BS, thus reducing the management cost for the BS. The shared key between two MSs is computed without any interaction between them. In addition, the trust among MSs is derived from the cellular infrastructure, thus enjoying an equal security level as provided in the underlying cellular network. Simulations are done to observe the system performance and the results are very encouraging.

Currently, inspection of remote environments poses potential hazards to the human operator. Mixed reality presents the unique opportunity for the human operators & the piloted robots to co-exist within a virtual world that is then spatially mapped to the physical world. This can then act as a platform to conduct remote inspection, where the human operators could potentially possess heightened situational awareness and control.