Integrating traffic cameras with deep learning facilitates real-time multi-camera vehicle path reconstruction, benefiting smart city initiatives such as traffic management and public safety. The high demands on latency, bandwidth, and privacy underscore the necessity of edge computing platforms. However, the continuous operation of compute-intensive deep learning algorithms on round-the-clock camera streams can exceed the compute and power capacities of edge data centers. Therefore, it is critical to strategically deploy and activate camera streams to balance compute demands with performance requirements. Currently, there is a lack of tools to support this planning process. This paper presents Seer, a comprehensive suite of tools and algorithms that aids traffic planners in optimizing traffic camera placement and deep learning model deployment, considering compute and budgetary constraints. Seer encompasses (1) a graph-based algorithm for efficient camera placement incorporating road network topology, (2) a collaborative algorithm for vehicle path reconstruction that enables the use of less accurate but resource-efficient deep learning models, and (3) a scalable traffic simulation framework derived from open-source projects that model city road networks. Evaluations of Seer on two real-world road networks, each with thousands of intersections and simulated vehicle paths, show a 6.3x reduction in compute requirements. This is achieved through a combination of sparse camera deployment and the use of lightweight models, with an average 55% increase in the Hausdorff distance for the path reconstruction algorithm, translating to an absolute increase of 135 to 195 meters.

Recent advancements in artificial intelligence (AI) have seen the emergence of smart video surveillance (SVS) in many practical applications, particularly for building safer and more secure communities in our urban environments. Cognitive tasks, such as identifying objects, recognizing actions, and detecting anomalous behaviors, can produce data capable of providing valuable insights to the community through statistical and analytical tools. However, artificially intelligent surveillance systems design requires special considerations for ethical challenges and concerns. The use and storage of personally identifiable information (PII) commonly pose an increased risk to personal privacy. To address these issues, this paper identifies the privacy concerns and requirements needed to address when designing AI-enabled smart video surveillance. Further, we propose the first end-to-end AI-enabled privacy-preserving smart video surveillance system that holistically combines computer vision analytics, statistical data analytics, cloud-native services, and end-user applications. Finally, we propose quantitative and qualitative metrics to evaluate intelligent video surveillance systems. The system shows the 17.8 frame-per-second (FPS) processing in extreme video scenes. However, considering privacy in designing such a system results in preferring the pose-based algorithm to the pixel-based one. This choice resulted in dropping accuracy in both action and anomaly detection tasks. The results drop from 97.48% to 73.72% in anomaly detection and 96% to 83.07% in the action detection task. On average, the latency of the end-to-end system is 36.1 seconds.

The falling cost of IoT cameras, the advancement of AI-based computer vision algorithms, and powerful hardware accelerators for deep learning have enabled the widespread deployment of surveillance cameras with the ability to automatically analyze streaming video feeds to detect events of interest. While streaming video analytics is currently largely performed in the cloud, edge computing has emerged as a pivotal component due to its advantages of low latency, reduced bandwidth, and enhanced privacy. However, a distinct gap persists between state-of-the-art computer vision algorithms and the successful practical implementation of edge-based streaming video analytics systems. This paper presents a comprehensive review of more than 30 research papers published over the last 6 years on IoT edge streaming video analytics (IE-SVA) systems. The papers are analyzed across 17 distinct dimensions. Unlike prior reviews, we examine each system holistically, identifying their strengths and weaknesses in diverse implementations. Our findings suggest that certain critical topics necessary for the practical realization of IE-SVA systems are not sufficiently addressed in current research. Based on these observations, we propose research trajectories across short-, medium-, and long-term horizons. Additionally, we explore trending topics in other computing areas that can significantly impact the evolution of IE-SVA systems.

The rise of AI-powered computer vision algorithms offers the possibility of visual sensing of the environment in IoT applications through the widespread use of low-cost video cameras. The need for low latency, bandwidth limitations, and privacy concerns associated with video data motivates the use of edge computing for computer vision applications. However, unlike cloud computing with almost unbounded resources, the edge is characterized by compute nodes of limited capacity and power budget. Additionally, fault tolerance is limited due to replication costs at the edge.In this poster, we present our initial work on evaluating the performance of an edge-specific version of Kubernetes on a Raspberry Pi4 cluster for multi-camera computer vision applications. Kubernetes enables automated deployment and management of containerized distributed applications to run at scale across a cluster of compute and storage nodes, while handling node failures. However, existing literature has not characterized the resource consumption and latency impact of Kubernetes for computer vision applications on realistic edge clusters. Our experimental results indicate that while Kubernetes can deliver fault tolerance at the edge, the choices made in the design of containers pods significantly affects the observed tail latency on a low power edge cluster.

The large scale use of real-time computer vision for IoT applications faces challenges of big data streams, complex processing, low latency requirements, and data privacy concerns. Edge computing allows data to be processed close to the source, vastly reducing the data that needs to be sent to the cloud, thus reducing network bandwidth requirements, and lowering application latency. Additionally, sensitive video streams can be confined to the privacy perimeter of the end-user. However, current IoT edge middleware are designed for low data rate sensor applications, and do not satisfy the demanding needs of computer vision-based IoT. In this paper, we present the design and implementation of a novel edge gateway targeted specifically at emerging IoT computer vision applications. The proposed edge gateway enables realization of multiple vision algorithms at the edge from a single camera stream. Furthermore, unlike existing edge gateways available from public cloud service providers, the proposed gateway is vendor-neutral, and capable of connecting to multiple cloud providers. This allows for increased application resilience, lower costs, and avoids cloud vendor lock-in. We experimentally evaluate the performance of the proposed edge gateway for multiple computer vision applications, and multiple public clouds.

Recently, Smart Video Surveillance (SVS) systems have been receiving more attention among scholars and developers as a substitute for the current passive surveillance systems. These systems are used to make the policing and monitoring systems more efficient and improve public safety. However, the nature of these systems in monitoring the public's daily activities brings different ethical challenges. There are different approaches for addressing privacy issues in implementing the SVS. In this paper, we are focusing on the role of design considering ethical and privacy challenges in SVS. Reviewing four policy protection regulations that generate an overview of best practices for privacy protection, we argue that ethical and privacy concerns could be addressed through four lenses: algorithm, system, model, and data. As an case study, we describe our proposed system and illustrate how our system can create a baseline for designing a privacy perseverance system to deliver safety to society. We used several Artificial Intelligence algorithms, such as object detection, single and multi camera re-identification, action recognition, and anomaly detection, to provide a basic functional system. We also use cloud-native services to implement a smartphone application in order to deliver the outputs to the end users.

Mez is a novel publish-subscribe messaging system for latency sensitive multi-camera machine vision applications at the IoT Edge. The unlicensed wireless communication in IoT Edge systems are characterized by large latency variations due to intermittent channel interference. To achieve user specified latency in the presence of wireless channel interference, Mez takes advantage of the ability of machine vision applications to temporarily tolerate lower quality video frames if overall application accuracy is not too adversely affected. Control knobs that involve lossy image transformation techniques that modify the frame size, and thereby the video frame transfer latency, are identified. Mez implements a network latency feedback controller that adapts to channel conditions by dynamically adjusting the video frame quality using the image transformation control knobs, so as to simultaneously satisfy latency and application accuracy requirements. Additionally, Mez uses an application domain specific design of the storage layer to provide low latency operations. Experimental evaluation on an IoT Edge testbed with a pedestrian detection machine vision application indicates that Mez is able to tolerate latency variations of up to 10x with a worst-case reduction of 4.2% of the application accuracy F1 score metric. The performance of Mez is also experimentally evaluated against state-of-the-art low latency NATS messaging system.

Machine vision applications at the IoT Edge have bandwdith and latency constraints due to large sizes of video data. In this paper we propose approximate computing, that trades off inference accuracy with video frame size, as a potential solution. We present a number of low compute overhead video frame modifications that can reduce the video frame size, while achieving acceptable levels of inference accuracy. We present, a heuristic based design space pruning, and a Categorical boost based machine learning model as two approaches to achieve scalable performance in determining the appropriate video frame modifications that satisfy design constraints. Experimental results on an object detection application on the Microsoft COCO 2017 data set, indicates that proposed methods were able to reduce the video frame size by upto 71.3% while achieving an inference accuracy of 80.9% of that of the unmodified video frames. The machine learning model has a high training cost, but has a lower inference time, and is scalable and flexible compared to the heuristic design space pruning algorithm.

Software bugs in a production environment have an undesirable impact on quality of service, unplanned system downtime, and disruption in good customer experience, resulting in loss of revenue and reputation. Existing approaches to automated software bug repair focuses on known bug templates detected using static code analysis tools and test suites, and in automatic generation of patch code for these bugs. We describe the typical bug fixing process employed in the Linux kernel, and motivate the need for a new automated tool flow to fix bugs. We present an initial design of such an automated tool that uses Recurrent Neural Network (RNN) based Natural Language Processing to generate patch recommendations from user generated bug reports. At the 50th percentile of the test bugs, the correct patch occurs within the top 11.5 patch recommendations output by the model. Further, we present a Linux kernel developer's assessment of the quality of patches recommended for new unresolved kernel bugs.

Recent advances in Deep Learning, have made possible distributed multi-camera vision analytics targeted at a variety of surveillance applications involving automated real-time analysis of events from multiple video perspectives. However, the latency critical nature of these applications necessitates computing at the Edge of the network, close to the cameras [1] . The required Edge computing infrastructure is necessarily distributed, with cloud like capabilities such as fault tolerance, scalability, multi-application tenancy, and security, while functioning at the unique operating environment of the Edge. Characteristics of the Edge include, highly heterogeneous hardware platforms with limited computational resources, variable latency wireless networks, and minimal physical security. To enable vision analytics at the Edge, application developers need a distributed middleware layer that provides a suitable abstraction of the Edge computing system, allowing cloud like DevOps workflow at the Edge. Middleware layers facilitate application development by providing suitable system abstractions thereby allowing the application programmers to focus on the applications needs rather than system details. In this poster, we present the requirements of such a middleware system. We propose a distributed messaging system with storage capabilities as a potential candidate for the middleware layer.

Motivation deficits and reduced goal-directed behaviour are prominent in schizophrenia (SZ), and significantly contribute to poor functional and treatment outcomes. One of the critical components of the multi-faceted motivation system is effort valuation, which refers to the mental processes involved in computing how much effort one is willing to exert in order to obtain a desired outcome. These effort-cost computations are typically measured using effort-based decision-making (EBDM) paradigms, where individuals must choose between performing low- or high-effort tasks for varying reward magnitudes. Rather than demonstrating an overall unwillingness to expend effort, however, studies have shown that individuals with SZ inefficiently allocate effort across different probability and reward conditions. Thus, in order to better understand the underlying computations involved in effort-based decision-making, the present study sought to model the predictors of choice behaviour in SZ and healthy control (HC) participants. Fifty-one SZ patients and 51 demographically-matched HC participants completed the Effort Expenditure for Rewards Task (EEfRT) as a measure of EBDM. In addition, all participants underwent characterization of clinical amotivation severity and cognitive functioning using the Apathy Evaluation Scale (AES) and Brief Assessment of Cognition in Schizophrenia (BACS), respectively. Generalized Estimating Equations (GEE) were subsequently applied to the EEfRT data with a binary logistic distribution used to model the likelihood of choosing hard tasks. A number of models were tested with independent variables including reward magnitude, probability, expected value (EV), diagnostic group, AES, and BACS. GEE models revealed significant main effects for reward magnitude (b = .54, p < .001), probability (b = .02, p < .001), and EV (b = .46, p < .001), but no main effect of group. However, significant interaction terms were found between group and reward (b = -.33, p < .001), group and probability (b = -.01, p = .007), and group and EV (b = -.58, p = .001). While there were no AES or BACS main effects, there were significant AES x reward (b = -.02, p < .001) and AES x EV (b = -.02, p = .01) interactions, as well as BACS x reward (b = .11, p < .001), BACS x probability (b = .01, p < .001), and BACS x EV (b = .31, p < .001) interactions. While SZ and HC participants are similarly willing to exert effort in pursuit of a reward, patients with SZ are less likely to utilize important information regarding the magnitude, probability, and expected value associated with that reward in driving their effort-based decision-making. Moreover, reward magnitude and EV are less predictive of effortful choices for individuals with greater motivation and cognitive impairments, regardless of their diagnostic status. Taken together, these findings suggest a direct link between amotivation, cognition, and inefficient utilization of reward and probability information in the context of choice behaviour and effort-cost computations.

Multicamera based Deep Learning vision applications subscribe to the Edge computing paradigm due to stringent latency requirements. However, guaranteeing latency in the wireless communication links between the cameras nodes and the Edge server is challenging, especially in the cheap and easily available unlicensed bands due to the interference from other camera nodes in the system, and from external sources. In this paper, we show how approximate computation techniques can be used to design a latency controller that uses multiple video frame image quality control knobs to simultaneously satisfy latency and accuracy requirements for machine vision applications involving object detection, and human pose estimation. Our experimental results on an Edge test bed indicate that the controller is able to correct for up to 164% degradation in latency due to interference within a settling time of under 1.15 s.

Motivation deficits have been linked to poor functional outcomes in schizophrenia (SZ) and major depressive disorder (MDD) and represent an unmet therapeutic need. Recent conceptualizations of the motivation system have outlined five inter-related reward processes, whereby (1) reward responsiveness (i.e. “liking”) and (2) reward prediction (i.e. “wanting”, established through appropriate reward learning) converge to inform both (3) reward valuation and (4) effort valuation which is associated with a cost-benefit analysis, followed by (5) the development and execution of an action plan to achieve the desired outcome. The inclusion of approach motivation within the RDoC Positive Valence System further underscores its importance as a pervasive symptom that cuts across traditional diagnostic boundaries. Previous investigations, however, have typically only focused on isolated reward processes, often within single diagnostic groups. Thus, in line with emerging dimensional approaches to examining psychopathology, the present study sought to utilize a cluster-analytic approach to objectively evaluate the multiple facets of the motivation system concurrently across SZ, MDD, and healthy control (HC) participants. The study sample consisted of 39 SZ, 38 MDD, and 39 HC participants. Participants were administered a series of assessments to evaluate symptom severity and cognitive functioning. Discrete facets of the motivation system were measured using an extensive battery of objective computerized tasks. Variables of interest were extracted from each task and entered into a principal components analysis in order to explore the factor structure of the motivation framework. Factor scores were subsequently applied to K-means cluster analysis to identify subgroups of individuals with similar motivation profiles. Principal components analysis revealed five distinct motivation factors: hedonic capacity, reward expectancy and learning, cost-benefit decision-making, goal-directed decision-making, and effort expenditure. K-means clustering identified two distinct subgroups of individuals based on their motivation task performance. The first cluster demonstrated impaired hedonic capacity (t(114)=-3.7, p<.001), whereas the second cluster was characterized by impairments in cost-benefit decision-making (t(114)=5.9, p<.001), goal-directed decision-making (t(114)=7.3, p<.001), and effort expenditure (t(114)=6.3, p<.001). Although clusters did not differ in symptom severity, the second cluster was associated with significantly greater cognitive impairments (t(114)=6.4, p<.001). Importantly, all diagnostic groups were well represented in each cluster, though with significantly different distributions. Our dimensional investigation revealed a multi-faceted motivation framework comprised of five distinct components. The emergence of two unique motivation performance profiles highlights the extensive heterogeneity of clinical amotivation and its dimensionality across disorders. Further, the pattern of motivation impairments within these profiles raises the possibility of distinct underlying neural substrates, with implications for specific therapeutic targets.

Recent advances in machine learning and deep learning have enabled many existing applications in smart cities, autonomous systems, and wearable devices. These applications often demand scalable real-time cognitive intelligence and on-the-spot decision making. Current computer systems have been customized for a cloud computing paradigm which often does not meet latency constraints and scalability requirements. To address the limitations of the cloud computing paradigm, the general trend is toward shifting the computation next to data producers at the edge. However, the edge computing paradigm is in the very early stages. Many system-level aspects of edge computing, including algorithms mapping and partitioning across edge computing resources (edge server, and edge nodes) are unknown. New research is required to understand and quantify design dimensions for edge computing.This paper presents a novel edge computing infrastructure for distributed real-time video analytics. This paper presents a holistic solution for co-designing application and edge infrastructure, including edge nodes and edge servers, to enable scalable real-time Artificial Intelligence (AI)/Deep Learning (DL) video analytics across many cameras. For experimental results and evaluation, we focus on the case study of object re-identification across many cameras, which is composed of object detection/classification (TinyYOLOv3), feature extraction, local re-identification, and global re-identification kernels. We evaluate the edge system under three different task mapping and resource allocation configurations. The results present that with the edge nodes (video cameras) more than 32, the only scalable solution is to perform detection/classification (TinyYOLOv3), feature extraction, local re-identification on the edge nodes next to cameras, and execute global re-identification on edge server.

Data Representation in memory is a one of task in Big Data. Data structures includes several types of tree data structures through system can access accurate and efficiently data in Big Data. Succinct data structures can play important role in data representation while data is processed in RAM memory for Big Data. Choosing a data structure for memory representation is very difficult problem in Big Data. We proposed some solution of problems of data representation in Big Data. Data Mining can be utilized to take decision by Data processing. We know the functions and rules for query processing. We have to either change method of data processing or we can change way of data representation in memory. In this paper, different kind of tree data structures are presented for data representation in RAM of computer system for Big Data by using succinct data structures. Data mining is often required in Big Data. Data must be processed in parallel or steaming manner. In this paper we first compare all data structures by table and then we proposed succinct data structures those are very popular now. Each tree presented for Data representation has different time and space complexities.

Postpartum depression (PPD) is the most common complication of childbearing with a 13% prevalence rate, but there are no widespread prevention strategies and no nutraceutical interventions have been developed. Postpartum blues (PPB) is often a prodromal state for PPD, since severe PPB strongly elevates risk for PPD. A dietary supplement kit consisting of monoamine precursor amino acids, tryptophan and tyrosine, and dietary antioxidants was created. The aim of this open-label study was to assess whether the dietary supplement reduces the vulnerability to depressed mood at day-5 postpartum, the typical peak of PPB.

Autonomous machine vision is a powerful tool to address challenges in multiple domains including national security (for example, video surveillance), health care (for example, patient monitoring), and transportation (for example, autonomous vehicles). Distributed vision, where multiple cameras observe a specific geographic area 24/7, enables smart understanding of events in a physical environment with minimal human intervention. We observe that the cloud paradigm alone does not offer a pathway to real-time distributed vision processing. With potentially thousands of cameras, hundreds of gigabytes data per second needs to be transferred to the cloud, saturating the bandwidth of the network. More importantly, vision applications are inherently latency-critical with a high demand for real-time scene analysis (for example, feature extraction and object tracking). To meet latency requirements, computation - including both processing of raw video streams to identify objects, and analytics on this data, needs to be brought to the edge of the network. While object recognition may be done locally at the end node (next to the camera), vision analytics requires access to data generated across different nodes. For example, a subject of interest may need to be tracked across multiple cameras to identify the nature of activities. This creates a need for a low latency distributed data store communicating over a dynamic communication network (most often wireless), to be implemented at the edge. Moreover, the data store must be able to address the limited storage at the end nodes (typically gigabytes). Additionally, privacy and security are prime concerns in the design of such a distributed edge storage.

Theory of mind-the ability to decode and reason about others' mental states-is a universal human skill and forms the basis of social cognition. Theory of mind accuracy is impaired in clinical conditions evidencing social impairment, including major depressive disorder. The current study is a preliminary investigation of the association of polymorphisms of the serotonin transporter (SLC6A4), dopamine transporter (DAT1), dopamine receptor D4 (DRD4), and catechol-O-methyl transferase (COMT) genes with theory of mind decoding in a sample of adults with major depression. Ninety-six young adults (38 depressed, 58 non-depressed) completed the 'Reading the Mind in the Eyes task' and a non-mentalistic control task. Genetic associations were only found for the depressed group. Specifically, superior accuracy in decoding mental states of a positive valence was seen in those homozygous for the long allele of the serotonin transporter gene, 9-allele carriers of DAT1, and long-allele carriers of DRD4. In contrast, superior accuracy in decoding mental states of a negative valence was seen in short-allele carriers of the serotonin transporter gene and 10/10 homozygotes of DAT1. Results are discussed in terms of their implications for integrating social cognitive and neurobiological models of etiology in major depression.