Video anomaly detection in real-time on a Power-Aware Heterogeneous Platform

FPGAs and GPUs are often used when real-time performance in video processing is required. An accelerated processor is chosen based on task-specific priorities (power consumption, processing time and detection accuracy), and this decision is normally made once at design time. All three characteristics are important, particularly in battery-powered systems. Here we propose a method for moving selection of processing platform from a single design-time choice to a continuous run time one.We implement Histogram of Oriented Gradients (HOG) detectors for cars and people and Mixture of Gaussians (MoG) motion detectors running across FPGA, GPU and CPU in a heterogeneous system. We use this to detect illegally parked vehicles in urban scenes. Power, time and accuracy information for each detector is characterised. An anomaly measure is assigned to each detected object based on its trajectory and location, when compared to learned contextual movement patterns. This drives processor and implementation selection, so that scenes with high behavioural anomalies are processed with faster but more power hungry implementations, but routine or static time periods are processed with power-optimised, less accurate, slower versions. Real-time performance is evaluated on video datasets including i-LIDS. Compared to power-optimised static selection, automatic dynamic implementation mapping is 10% more accurate but draws 12W extra power in our testbed desktop system.

Person Re-Identification using Deep Convnets with Multi-task Learning

Person re-identification involves recognizing a person across non-overlapping camera views, with different pose, illumination, and camera characteristics. We propose to tackle this problem by training a deep convolutional network to represent a person’s appearance as a low-dimensional feature vector that is invariant to common appearance variations encountered in the re-identification problem. Specifically, a Siamese-network architecture is used to train a feature extraction network using pairs of similar and dissimilar images. We show that use of a novel multi-task learning objective is crucial for regularizing the network parameters in order to prevent over-fitting due to the small size the training dataset. We complement the verification task, which is at the heart of re-identification, by training the network to jointly perform verification, identification, and to recognise attributes related to the clothing and pose of the person in each image. Additionally, we show that our proposed approach performs well even in the challenging cross-dataset scenario, which may better reflect real-world expected performance. 

GPs’ and community pharmacists’ opinions on medication management at transitions of care in Ireland

Objective. The aim of this study was to survey GPs and community pharmacists (CPs) in Ireland regarding current practices of medication management, specifically medication reconciliation, communication between health care providers and medication errors as patients transition in care.
Methods. A national cross-sectional survey was distributed electronically to 2364 GPs, 311 GP Registrars and 2382 CPs. Multivariable associations comparing GPs to CPs were generated and content analysis of free text responses was undertaken.
Results. There was an overall response rate of 17.7% (897 respondents—554 GPs/Registrars and 343 CPs). More than 90% of GPs and CPs were positive about the effects of medication reconciliation on medication safety and adherence. Sixty per cent of GPs reported having no formal system of medication reconciliation. Communication between GPs and CPs was identified as good/very good by >90% of GPs and CPs. The majority (>80%) of both groups could clearly recall prescribing errors, following a transition of care, they had witnessed in the previous 6 months. Free text content analysis corroborated the positive relationship between GPs and CPs, a frustration with secondary care communication, with many examples given of prescribing errors.
Conclusions. While there is enthusiasm for the benefits of medication reconciliation there are limited formal structures in primary care to support it. Challenges in relation to systems that support inter-professional communication and reduce medication errors are features of the primary/secondary care transition. There is a need for an improved medication management system. Future research should focus on the identified barriers in implementing medication reconciliation and systems that can improve it.

Cooperative Mapping and Exploration using Counter-rotational Potential Fields

This paper presents a novel technique for mapping and exploration using cooperating autonomous underwater vehicles. Rather than using the typical lawnmower sweep pattern to search an entire area, the proposed navigational plan involves guiding the formation directly towards each object of interest in turn, before arriving at a final goal position. This is achieved by the use of traditional artificial potential fields alongside counter-rotational potential fields. These clockwise and counter-clockwise fields are employed simultaneously by vehicles to ensure that the entire object is scanned rather than simply avoided as is the case with traditional collision avoidance techniques. The proposed methodology allows a formation to have fluid-like motion whilst a separation distance between cooperating agents (free of angular constraints) is maintained with a greater degree of flexibility than traditional formation control approaches. Owing to its nature, this technique is suited for applications such as exploration, mapping and underwater inspection to name a few. Simulation results demonstrate the efficacy of the proposed approach.