Sound Can Help Us See More Clearly

In the field of video action classification, existing network frameworks often only use video frames as input. When the object involved in the action does not appear in a prominent position in the video frame, the network cannot accurately classify it. We introduce a new neural network structure that uses sound to assist in processing such tasks. The original sound wave is converted into sound texture as the input of the network. Furthermore, in order to use the rich modal information (images and sound) in the video, we designed and used a two-stream frame. In this work, we assume that sound data can be used to solve motion recognition tasks. To demonstrate this, we designed a neural network based on sound texture to perform video action classification tasks. Then, we fuse this network with a deep neural network that uses continuous video frames to construct a two-stream network, which is called A-IN. Finally, in the kinetics dataset, we use our proposed A-IN to compare with the image-only network. The experimental results show that the recognition accuracy of the two-stream neural network model with uesed sound data features is increased by 7.6% compared with the network using video frames. This proves that the rational use of the rich information in the video can improve the classification effect.

BSIRNet: A Road Extraction Network with Bidirectional Spatial Information Reasoning

Automatic extraction of road information from remote sensing images is widely used in many fields, such as urban planning and automatic navigation. However, due to interference from noise and occlusion, the existing road extraction methods can easily lead to road discontinuity. To solve this problem, a road extraction network with bidirectional spatial information reasoning (BSIRNet) is proposed, in which neighbourhood feature fusion is used to capture spatial context dependencies and expand the receptive field, and an information processing unit with a recurrent neural network structure is used to capture channel dependencies. BSIRNet enhances the connectivity of road information through spatial information reasoning. Using the public Massachusetts road dataset and Wuhan University road dataset, the superiority of the proposed method is verified by comparing its results with those of other models.

Ship Type Recognition Based on Ship Navigating Trajectory and Convolutional Neural Network

With the aim to solve the problem of missing or tampering of ship type information in AIS information, in this paper, a novel ship type recognition scheme based on ship navigating trajectory and convolutional neural network (CNN) is proposed. Firstly, according to speed and acceleration of the ship, three ship navigating situations, i.e., static, normal navigation and maneuvering, are integrated into the process of trajectory images generation in the form of pixels. Then, three kinds of modular network structures with different depths are trained and optimized to determine the appropriate convolutional neural network structure. In the validation phase of the model, a large amount of verified data with a time span of one month was used, covering a variety of water conditions including open water, ports, rivers and lakes. Following this approach, a kind of CNN scheme which can be directly used to identify ship types in a wide range of waters is proposed. This scheme can be used to judge the ship type when the static information is completely missing and to test the data when the ship type information is partially missing.

Neuroevolutionary reinforcing learning of neural networks

The article presents the results of combining 4 different types of neural network learning: evolutionary, reinforcing, deep and extrapolating. The last two are used as the primary method for reducing the dimension of the input signal of the system and simplifying the process of its training in terms of computational complexity.In the presented work, the neural network structure of the control device of the modeled system is formed in the course of the evolutionary process, taking into account the currently known structural and developmental features of self-learning systems that take place in living nature. This method of constructing it makes it possible to bypass the specific limitations of models created on the basis of recombination of already known topologies of neural networks.

SMYOLO: Lightweight Pedestrian Target Detection Algorithm in Low-Altitude Scenarios

Due to the arbitrariness of the drone’s shooting angle of view and camera movement and the limited computing power of the drone platform, pedestrian detection in the drone scene poses a greater challenge. This paper proposes a new convolutional neural network structure, SMYOLO, which achieves the balance of accuracy and speed from three aspects: (1) By combining deep separable convolution and point convolution and replacing the activation function, the calculation amount and parameters of the original network are reduced; (2) by adding a batch normalization (BN) layer, SMYOLO accelerates the convergence and improves the generalization ability; and (3) through scale matching, reduces the feature loss of the original network. Compared with the original network model, SMYOLO reduces the accuracy of the model by only 4.36%, the model size is reduced by 76.90%, the inference speed is increased by 43.29%, and the detection target is accelerated by 33.33%, achieving minimization of the network model volume while ensuring the detection accuracy of the model.

Deep Learning in the Classification of Stage of Liver Fibrosis in Chronic Hepatitis B with Magnetic Resonance ADC Images

Liver fibrosis in chronic hepatitis B is the pathological repair response of the liver to chronic injury, which is a key step in the development of various chronic liver diseases to cirrhosis and an important link affecting the prognosis of chronic liver diseases. The further development of liver fibrosis in chronic hepatitis B can lead to the disorder of hepatic lobule structure, nodular regeneration of hepatocytes, formation of a pseudolobular structure, namely, cirrhosis, clinical manifestations of liver dysfunction, and portal hypertension. So far, the diagnosis of liver fibrosis in chronic hepatitis B has been made manually by doctors. However, this is very subjective and boring for doctors. Doctors are likely to be interfered with by external factors, such as fatigue and lack of sleep. This paper proposed a 5-layer deep convolution neural network structure for the automatic classification of liver fibrosis in chronic hepatitis B. In the 5-layer deep convolution neural network structure, there were three convolution layers and two fully connected layers, and each convolution layer was connected with a pooling layer. 123 ADC images were collected, and the following results were obtained: the accuracy, sensitivity, specificity, precision, F1, MCC, and FMI were 88.13% ± 1.47%, 81.45% ± 3.69%, 91.12% ± 1.72%, 80.49% ± 2.94%, 80.90% ± 2.39%, 72.36% ± 3.39%, and 80.94% ± 2.37%, respectively.

Application of the Kohonen neural network for monitoring tissue oxygen supply under hypoxic conditions

The article presents the results of application of the Kohonen artificial neural network (KANN) in assessing the oxygen status of human tissues, as well as for studying the adaptive-compensatory response of the body to functional load. In the experiment, the registered digital oxygen images of the tissue of 31 subjects were distributed into three classes using the KANN. Each group is characterised by different resistance of the organism to hypoxia. The research results have shown the effectiveness of using an artificial neural network structure and the possibility of its implementation for recognition of the functional state of a person under conditions of metabolic hypoxia; it seems relevant and has theoretical and practical significance in the framework of ecological physiology.

Method of creation of a two-level neural network structure for solving problems in mechanical engineering

There is an industrial trend of increasing technological effectiveness of production and growing autonomy of factories, resulting in generation of huge amounts of raw data that can be used to improve efficiency and continuity of industrial plants operation. Big data collection, processing and analysis technologies are already being actively implemented. Further use of big data will help with a variety of tasks, such as optimization and process monitoring, quality control of equipment and produced parts, modeling and forecasting of the facility operation and other mechanical engineering challenges. A new method of creation of a two-level neural network structure is proposed, designed to solve a number of problems by training individual neural networks for each subset of data used in the task at hand. This method combines two levels of information processing: the first level of the neural network classifier and the second level, which includes several neural network analyzers. Depending on the specific subject area and the data sets available, it is possible to use the method to solve various problems in mechanical engineering. The method allows to add new neural network analyzers and expand the scope of application. The practical application of the method in solving the problem of text message sentiment analysis is shown and an example of the Python programming language software implementation of the two-level structure is given. Use cases for the two-level structure method in mechanical engineering tasks are proposed. In addition, the proposed method can be used as a part of the hybrid intelligent information system that includes mivar expert systems. Combining neural networks with mivar expert systems as part of a hybrid intelligent information system is a promising direction for the development of artificial intelligence for mechanical engineering.