As the Internet-of-Things (IoT) and edge computing have been major paradigms for distributed data collection, communication, and processing, smart city applications in the real world tend to adopt IoT and edge computing broadly. Today, more and more machine learning algorithms would be deployed into front-end sensors, devices, and edge data centres rather than centralised cloud data centres. However, front-end sensors and devices are usually not so capable as those computing units in huge data centres, and for this sake, in practice, engineers choose to compromise for limited capacity of embedded computing and limited memory, e.g., neural network models being pruned to fit embedded devices. Visual object tracking is one of many important elements of a smart city, and in the IoT and edge computing context, high requirements to computing power and memory space severely prevent massive and accurate tracking. In this paper, we report on our contribution to object tracking on lightweight computing including (1) using limited computing capacity and memory space to realise tracking; (2) proposing a new algorithm region proposal correlation filter fitting for most edge devices. Systematic evaluations show that (1) our techniques can fit most IoT devices; (2) our techniques can keep relatively high accuracy; and (3) the generated model size is much less than others.
5 January 2021 in ACM Transactions on Internet Technology
Urban crime is an ongoing problem in metropolitan development and attracts general concern from the international community. As an effective means of defending urban safety, crime prediction plays a crucial role in patrol force allocation and public safety. However, urban crime data is a macro result of crime patterns overlapped by various irrelevant factors that cause inhomogeneous noises—local outliers and irregular waves. These noises might obstruct the learning process of crime prediction models and result in a deviation of performance. To tackle the problem, we propose a novel paradigm of Dual-robust Enhanced Spatial-temporal Learning Network (DuroNet), an encoder-decoder architecture that possesses an adaptive robustness for reducing the effect of outliers and waves. The robustness is mainly reflected on two aspects. One is a locality enhanced module that employs local temporal context information to smooth the deviation of outliers and dynamic spatial information to assist in understanding normal points. The other is a self-attention-based pattern representation module to weaken the effect of irregular waves by learning attentive weights. Finally, extensive experiments are conducted on two real-world crime datasets before and after adding Gaussian noises. The results demonstrate the superior performance of our DuroNet over the state-of-the-art methods.
1 October 2019 in Journal of Parallel and Distributed Computing
The emergence of new wearable technologies, such as action cameras and smart glasses, has driven the use of the first-person perspective in computer applications. This field is now attracting the attention and investment of researchers aiming to develop methods to process first-person vision (FPV) video. The current approaches present particular combinations of different image features and quantitative methods to accomplish specific objectives, such as object detection, activity recognition, user–machine interaction, etc. FPV-based navigation is necessary in some special areas, where Global Position System (GPS) or other radio-wave strength methods are blocked, and is especially helpful for visually impaired people. In this paper, we propose a hybrid structure with a convolutional neural network (CNN) and local image features to achieve FPV pedestrian navigation. A novel end-to-end trainable global pooling operator, called AlphaMEX, has been designed to improve the scene classification accuracy of CNNs. A scale-invariant feature transform (SIFT)-based tracking algorithm is employed for movement estimation and trajectory tracking of the person through each frame of FPV images. Experimental results demonstrate the effectiveness of the proposed method. The top-1 error rate of the proposed AlphaMEX-ResNet outperforms the original ResNet (k = 12) by 1.7% on the ImageNet dataset. The CNN-SIFT hybrid pedestrian navigation system reaches 0.57 m average absolute error, which is an adequate accuracy for pedestrian navigation. Both positions and movements can be well estimated by the proposed pedestrian navigation algorithm with a single wearable camera.
18 September 2020 in IEEE Transactions on Sustainable Computing
Dependence of computing resources on each other in cloud computing systems (CCS) makes them prone to fail in correlated manner which significantly impacts their service reliability and energy efficiency. Focusing on these two metrics of CCS while considering correlated failures remained an open question, which is the focus of this work. This paper proposes mechanisms for improving reliability and energy efficiency jointly under correlated failures in CCS. In order to model failure correlation, statistical cluster analysis techniques are applied to real failure traces. Then, mathematical models are built to calculate reliability and energy consumption of failure prone CCS. These mathematical models are used to design fault-tolerant and energy-aware resource provisioning mechanisms/policies. In order to further reduce the energy consumption, a correlated failure-aware VM consolidation policy is also proposed in this paper. A simulation based study of the proposed resource management policies and fault tolerance mechanisms is conducted by using real failure traces and Bag-of-Tasks workload. The results demonstrate that by exploiting failure correlation with the proposed resource management policies, we reduce the occurrence of failures on tasks by 34% and increase the energy efficiency of the system by 20%, approximately in comparison to the environments where failures are handled independently.