RoadCare: A Deep-learning Based Approach to Quantifying Road Surface Quality

Road traffic injuries.

Abstract: The appearance of cars has raised materialistic civilization and living standard to an unprecedented level. Today, it is hard to imagine how we human beings can live without cars. Yet, motor vehicles can cause a great number of deaths and injuries as well as considerable economic losses, which have constituted the global burden. Understanding of the occurrence and development of road traffic injuries will contribute to the prevention and control of crash and to the implementation of "everybody has the right to enjoy health" proposed by WHO.

Anomalies Detection Through Smartphone Sensors: A Review

Abstract: With smartphones being so ubiquitous, more connected and largely fitted with several types of sensors such as GPS, microphones, cameras, magnetometers, accelerometers, etc; there is an increasing opportunity in the development of smartphone-based sensor systems. In this article, we propose a short survey on the use of smartphone sensors in the detection of various kinds of anomalies in several fields namely environment, agriculture, healthcare and road/traffic conditions. We also list the main advantages and limitations of the use of smartphone sensors systems in such fields.

Machine Learning Approaches for Rapid Pothole Detection from 2D Images

Abstract: Roads are inevitable parts of human civilisation, and construction of roads are considered under a Civil Engineering problem; but periodically these roads require maintenance and assessment, which is highly dependent on adequate and timely pavement condition data. Howbeit, in some cases, it has been found that the manual practice of collecting and analysing such data often leads to delay in reporting about the issues and fixing them on time. Also, repairing potholes is time consuming, and locating these manually is a huge task. We want to find out some mechanism which can identify the construction conditions as well as any kind of deformities on the road from the dashboard camera fitted into a car, and at the same time, can analyse the conditions of road surface and formation of potholes on the road. Optimization of manual pothole detection through automation has been a part of scientific research since long. Pothole identification has significantly been adapted in different screening and maintenance systems. But in our country, owing to the large number of road networks and wide variations in the nature of rural and urban road conditions, it is very difficult to identify potholes through an automated system. In this paper, we have looked into several methods of Computer Vision, like image processing techniques and object detection method so as to identify potholes from the video input stream to the system. But these techniques have been found to have different challenges like lighting conditions, interference in the line of vision on waterlogged roads, and inefficiency at night vision. Hence, furthermore, we have explored the viability of Deep Learning method for identifying the potholes from the processing of input video streams, and have also analysed the Convolutional Neural Networks approach of Deep Learning through a self-built CNN model. In this paper, the expediency of all the methods as well as their drawbacks have been discussed.

Understanding Driver-Passenger Interactions in Vehicular Crowdsensing

Abstract: Smart city projects collect data on urban environments to identify problems, inform policymaking, and boost citizen engagement. Typically, this data is collected by static sensors placed around the city, which is not ideal for spatiotemporal needs of certain sensing applications such as air quality monitoring. Vehicular crowdsensing is an upcoming approach that addresses this problem by utilizing vehicles' mobility to collect fine-grained city-scale data. Prior work has mainly focused on designing vehicular crowdsensing systems and related components, including incentive schemes, vehicle selection, and application-specific sensing, without understanding the motivations and challenges faced by drivers and passengers, one of the two key stakeholders of any vehicular crowdsensing solution. Our work aims to fill this gap. To understand drivers' and passengers' perspectives, we developed Turn2Earn, a generic vehicular crowdsensing system that incentivizes drivers to take specific routes for data collection. Turn2Earn system was deployed with 13 auto-rickshaw drivers for two weeks in Bangalore, India. Our drivers took 709 trips using Turn2Earn covering 79.2% of the city's grid cells. Interviews with 13 drivers and 15 passengers revealed innovative information-based strategies adopted by the drivers to convince passengers in taking alternative routes, and passengers' altruism in supporting the drivers. We uncovered novel insights, including viability of offered routes due to road closure, issues with electric vehicles, and selection bias among the drivers. We conclude with design recommendations to inform the future of vehicular crowdsensing, including engaging and incentivizing passengers, and criticality-based reward structure.

ImageNet Classification with Deep Convolutional Neural Networks

Abstract: We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0% which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overriding in the fully-connected layers we employed a recently-developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Abstract: Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization, and in some cases eliminates the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.82% top-5 test error, exceeding the accuracy of human raters.

SURF: speeded up robust features

Abstract: In this paper, we present a novel scale- and rotation-invariant interest point detector and descriptor, coined SURF (Speeded Up Robust Features). It approximates or even outperforms previously proposed schemes with respect to repeatability, distinctiveness, and robustness, yet can be computed and compared much faster. This is achieved by relying on integral images for image convolutions; by building on the strengths of the leading existing detectors and descriptors (in casu, using a Hessian matrix-based measure for the detector, and a distribution-based descriptor); and by simplifying these methods to the essential. This leads to a combination of novel detection, description, and matching steps. The paper presents experimental results on a standard evaluation set, as well as on imagery obtained in the context of a real-life object recognition application. Both show SURF's strong performance.

Silhouettes: a graphical aid to the interpretation and validation of cluster analysis

Abstract: A new graphical display is proposed for partitioning techniques. Each cluster is represented by a so-called silhouette, which is based on the comparison of its tightness and separation. This silhouette shows which objects lie well within their cluster, and which ones are merely somewhere in between clusters. The entire clustering is displayed by combining the silhouettes into a single plot, allowing an appreciation of the relative quality of the clusters and an overview of the data configuration. The average silhouette width provides an evaluation of clustering validity, and might be used to select an ‘appropriate’ number of clusters.