A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects

Abstract: A convolutional neural network (CNN) is one of the most significant networks in the deep learning field. Since CNN made impressive achievements in many areas, including but not limited to computer vision and natural language processing, it attracted much attention from both industry and academia in the past few years. The existing reviews mainly focus on CNN's applications in different scenarios without considering CNN from a general perspective, and some novel ideas proposed recently are not covered. In this review, we aim to provide some novel ideas and prospects in this fast-growing field. Besides, not only 2-D convolution but also 1-D and multidimensional ones are involved. First, this review introduces the history of CNN. Second, we provide an overview of various convolutions. Third, some classic and advanced CNN models are introduced; especially those key points making them reach state-of-the-art results. Fourth, through experimental analysis, we draw some conclusions and provide several rules of thumb for functions and hyperparameter selection. Fifth, the applications of 1-D, 2-D, and multidimensional convolution are covered. Finally, some open issues and promising directions for CNN are discussed as guidelines for future work.

NeoRL: A Near Real-World Benchmark for Offline Reinforcement Learning

Abstract: Offline reinforcement learning (RL) aims at learning effective policies from historical data without extra environment interactions. During our experience of applying offline RL, we noticed that previous offline RL benchmarks commonly involve significant reality gaps, which we have identified include rich and overly exploratory datasets, degraded baseline, and missing policy validation. In many real-world situations, to ensure system safety, running an overly exploratory policy to collect various data is prohibited, thus only a narrow data distribution is available. The resulting policy is regarded as effective if it is better than the working behavior policy; the policy model can be deployed only if it has been well validated, rather than accomplished the training. In this paper, we present a Near real-world offline RL benchmark, named NeoRL, to reflect these properties. NeoRL datasets are collected with a more conservative strategy. Moreover, NeoRL contains the offline training and offline validation pipeline before the online test, corresponding to real-world situations. We then evaluate recent state-of-the-art offline RL algorithms in NeoRL. The empirical results demonstrate that some offline RL algorithms are less competitive to the behavior cloning and the deterministic behavior policy, implying that they could be less effective in real-world tasks than in the previous benchmarks. We also disclose that current offline policy evaluation methods could hardly select the best policy. We hope this work will shed some light on future research and deploying RL in real-world systems.

A Text-Based CAPTCHA Cracking System with Generative Adversarial Networks

Abstract: As a multimedia security mechanism, CAPTCHAs are completely automated public turing test to tell computers and humans apart. Although cracking CAPTCHA has been explored for many years, it is still a challenging problem for real practice. In this demo, we present a text based CAPTCHA cracking system by using convolutional neural networks(CNN). To solve small sample problem, we propose to combine conditional deep convolutional generative adversarial networks(cDCGAN) and CNN, which makes a tremendous progress in accuracy. In addition, we also select multiple models with low pearson correlation coefficients for majority voting ensemble, which further improves the accuracy. The experimental results show that the system has great advantages and provides a new mean for cracking CAPTCHAs.

A convolutional neural network based approach towards real-time hard hat detection

Abstract: Health and safety management has been an important issue in construction industry. National regulations impose the using of hard hats in construction sites. However, there are often cases in which the construction workers neglect the regulations. It is desired to monitor the correct wearing of hard hat in real time and explore monitoring techniques facilitated by deep-learning algorithms. In this paper, a convolutional neural network based hard-hat detection algorithm is proposed. In this algorithm, the detection of construction workers and the hard hats are assisted by computer vision technique where deep learning model are trained to identify the proper wearing of hard hats. The optimization of the proposed neural networks can reduce the computational complexity while maintaining a relatively high recognition precision. Experiments have been performed using five different algorithms for comparison and results demonstrate that the proposed algorithm excels in the mAP and FPS performance metrics. The experimental results collected on an embedded platform reveal that the proposed algorithm presents a good candidate for similar applications where real-time deep-learning application is desired.

Review of prediction of water-saving potentials

Abstract: The classification, principle, scope of application and research status of prediction methods for water-saving potentials in different fields are summarized so as to help researchers to select models more quickly for practical problems. The results show that the formula model-based methods have been widely used and are easy to operate with less data. However, their precisions are not high. The machine learning-based methods can be widely utilized with high precision although they require more data. In response to the existing problems in the prediction of water-saving potentials, the development trend is summarized and analyzed. In the future, some new technologies such as deep learning, big data will be introduced into the researches on the prediction of water-saving potentials according to different industries and regional characteristics. They will look forward to achieve refined water saving potential prediction. In addition, accelerating the application of Internet-based prediction of water-saving potential will realize the water-saving socialization services which integrate the functions such as data collection, processing, forecasting and publishing.

Face Image Restoration Based on Residual Generative Adversarial Network

Abstract: Benefiting from the rapid development of computer vision, face image restoration technology can only use the contour of the face to generate a complete face image. At present, many face restoration techniques based on convolutional neural networks and generative adversarial networks have been proposed. They can restore partial damaged face images or even directly generate face images using face contours. However, the results of qualitative and quantitative analysis of face images restored by these techniques are not ideal, and there are many limitations in the restoration process. Therefore, this paper proposes a face image restoration method based on residual generative adversarial network (FR-RGAN), which improves the performance of the model by means of deep convolution, residual network, and smaller convolution kernels, and restores the local details of the face by using the contour of the face, making it more vivid. Experimental results show that, compared with pix2pix, FR-RGAN has an improvement of 8.7%, 2.1%, and 9.6% respectively in mean square error, peak signal to noise ratio, and structural similarity index, and 53.4%, 12.6%, and 30.1% better than non-residual method.