With the rapid development of Internet of Things (IoT) devices and network infrastructure, there have been a lot of sensors adopted in the industrial productions, resulting in a large size of data. One of the most popular examples is the manufacture inspection, which is to detect the defects of the products. In the paper, in order to implement a robust inspection system with higher accuracy, we propose a deep learning based classification model, in order to find the possible defective products. As there may be many assembly lines in one factory, one huge problem in this scenario is how to process such big data in real time. Therefore, we design our system with the concept of fog computing. By offloading the computation burden from the central server to the fog nodes, the system obtains the ability to deal with extremely large data. There are two obvious advantages in our system. The first one is that we adapt the convolutional neural network (CNN) model into the fog computing environment, which significantly improves its computing efficiency. The other one is we work out an inspection model which can simultaneously indicate the defect type and its degree. The experimental results demonstrate the robustness and efficiency of the proposed method.

Retinal vessel segmentation is of great interest for diagnosis of retinal vascular diseases. To further improve the performance of vessel segmentation, we propose IterNet, a new model based on UNet [1], with the ability to find obscured details of the vessel from the segmented vessel image itself, rather than the raw input image. IterNet consists of multiple iterations of a mini-UNet, which can be 4× deeper than the common UNet. IterNet also adopts the weight-sharing and skip-connection features to facilitate training; therefore, even with such a large architecture, IterNet can still learn from merely 10~20 labeled images, without pre-training or any prior knowledge. IterNet achieves AUCs of 0.9816, 0.9851, and 0.9881 on three mainstream datasets, namely DRIVE, CHASE-DB1, and STARE, respectively, which currently are the best scores in the literature. The source code is available ¹ .

The autonomous vehicle, as an emerging and rapidly growing field, has received extensive attention for its futuristic driving experiences. Although the fast developing depth sensors and machine learning methods have given a huge boost to self-driving research, existing autonomous driving vehicles do meet with several avoidable accidents during their road testings. The major cause is the misunderstanding between self-driving systems and human drivers. To solve this problem, we propose a humanlike driving system in this paper to give autonomous vehicles the ability to make decisions like a human. In our method, a convolutional neural network model is used to detect, recognize, and abstract the information in the input road scene, which is captured by the on-board sensors. And then a decision-making system calculates the specific commands to control the vehicles based on the abstractions. The biggest advantage of our work is that we implement a decision-making system which can well adapt to real-life road conditions, in which a massive number of human drivers exist. In addition, we build our perception system with only the depth information, rather than the unstable RGB data. The experimental results give a good demonstration of the efficiency and robustness of the proposed method.

Electrical load forecasting is still a challenging open problem due to the complex and variable influences (e.g., weather and time). Although, with the recent development of IoT and smart meter technology, people have obtained the ability to record relevant information on a large scale, traditional methods struggle in analyzing such complicated relationships for their limited abilities in handling nonlinear data. In the article, we introduce an IoTbased deep learning system to automatically extract features from the captured data, and ultimately, give an accurate estimation of future load value. One significant advantage of our method is the specially designed two-step forecasting scheme, which significantly improves the forecasting precision. Also, the proposed method is able to quantitatively analyze the influences of some major factors, which is of great guiding significance to select attribute combination and deploy onboard sensors for smart grids with vast areas, variable climates, and social conventions. Simulations demonstrate that our method outperforms some existing approaches, and can be well applied in various situations.

Electrical load forecasting is of great significance to guarantee the system stability under large disturbances, and optimize the distribution of energy resources in the smart grid. Traditional prediction models, which are mainly based on time series analyzing, have been unable to fully meet the actual needs of the power system, due to their non-negligible prediction errors. To improve the forecasting precision, we successfully transform the numerical prediction problem into an image processing task, and, based on that, utilize the state-of-the-art deep learning methods, which have been widely used in computer image area, to perform the electrical load forecasting. A novel deep learning based short-term forecasting (DLSF) method is proposed in the paper. Our method can perform accurate clustering on the input data using a deep Convolutional Neural Network (CNN) model. And ultimately, another neural network with three hiddenlayers is used to predict the electric load, considering various external influencing factors, e.g. temperature, humidity, wind speed, etc. Experimental results demonstrate that the proposed DLSF method performs well in both accuracy and efficiency.

Traditional Network Functions Virtualization (NFV) implementations are somehow too heavy and do not have enough functionality to conduct complex tasks. In this work, we propose a light-weight NFV framework named DeepNFV, which is based on the Docker container running on the network edge, and integrates state-of-the-art deep learning models with NFV containers to address some complicated problems, such as traffic classification, link analysis, and so on. We compare the DeepNFV framework with several existing works, and detail its structures and functions. The most significant advantage of DeepNFV is its light-weight design, resulting from the virtualization and low-cost nature of the container technology. Also, we design this framework to be compatible with edge devices, in order to decrease the computational overhead of the central servers. Another merit is its strong analysis ability brought by deep learning models, which make it suitable for many more scenarios than traditional NFV approaches. In addition, we also describe some typical application scenarios, regarding how the NFV container works and how to utilize its learning ability. Simulations demonstrate its high efficiency, as well as the outstanding recognition performance in a typical use case.