Convolutional Attention Module Research Articles

A Hybrid Approach for the Detection and Classification of Tomato Leaf Diseases

In this paper, we proposed a hybrid deep learning approach for detecting and classifying tomato plant leaf diseases early. This hybrid system is a combination of a convolutional neural network (CNN), convolutional attention module (CBAM), and support vector machines (SVM). Initially, the proposed model can detect nine different tomato diseases but is not limited to this. The proposed system is tested using a database containing images of tomato leaves. The obtained results were very encouraging, giving us accuracy up to 97.2%, which can be improved with the improvement of learning processes. The proposed system is very efficient and lightweight, so the farmer can install it on any smart device having a digital camera and processing capabilities. With a bit of training, a farmer can detect any disease immediately, which will help him take timely pre-emptive action.

Evaluation of Tourism Food Safety and Quality with Neural Networks.

Food safety issues are inextricably linked to people's lives and, in extreme cases, endanger public safety and social stability. People are becoming increasingly concerned about food safety issues in a modern society with high-quality economic development. People's incomes are increasing day by day as the economy continues to grow, and the tourism industry has grown by leaps and bounds. However, many problems arose, such as the issue of food safety in tourism. Tourism food safety issues affect not only the development of the food industry but also the development of tourism. Food safety oversight of tourist attractions has always been a relatively concerning issue in the country, and it is also something that the general public is concerned about. It can be said that food safety supervision of tourist attractions is the most important thing in food safety supervision. In this context, it becomes an important task to evaluate the safety of tourist food. This work proposes a multiscale convolutional neural network (AMCNN) combined with neural networks and attention layers to realize the safety and quality evaluation of tourist food. The algorithm uses the lightweight Xception network as a basic model and utilizes multiscale depth-separable convolution modules of different sizes for feature extraction and fusion to extract richer food safety feature information. Furthermore, the convolutional attention module (CBAM) is embedded on the basis of the multiscale convolutional neural network, which makes the network model focus more on discriminative features.

CSI-Based Location-Independent Human Activity Recognition Using Feature Fusion

Channel state information (CSI)-based human activity recognition (HAR) has important application prospects, such as smart homes, medical monitoring, and public security. Due to the collected CSI data contains not only activity information but also activity-unrelated environmental information, the characteristics of the same activity conducted at different locations are different. The existing methods of collecting samples at a fixed location to train the HAR model can hardly work well at other locations, which limits the application prospect of CSI-based HAR. To deal with this challenge, we proposed an Attention-based feature Fusion ACTivity recognition system (AF-ACT). The proposed system extracts the semantic activity features and temporal features from different dimensions to better characterize the activity at different locations. The semantic activity features are extracted by the convolutional neural network (CNN) combined with the convolutional attention module (CBAM), and the temporal features are extracted by bidirectional gated recurrent unit (BGRU) combined with the self-attention mechanism. The semantic activity features and temporal features are fused through an attention-based feature fusion (A-Fusion) module to obtain complementary information, which will promote recognition accuracy. The proposed system is evaluated in an open environment with 12 activity training locations and ten arbitrary testing locations. The experimental results show that the system can reach the highest accuracy of 91.23% in different experimental conditions when recognizing eight categories of activities.

A Detection Mechanism for PMU Data Manipulation Attacks by Using Sliced Recurrent Convolutional Attention Module

Background: Phasor Measurement Unit (PMU) Data Manipulation Attacks (PDMA) can change the state estimates of power systems and cause significant damage to the smart grid. So it is vital to research a method to detect it. Objective: In this paper, we propose a detection mechanism and model for PDMA. Methods: Firstly, the distribution's characteristics of Phasor Data Concentrator (PDC) and PMU are analyzed, and we use these characteristics to detect a PDMA detection mechanism that could help us reduce the number of detection samples. Secondly, we use the Sliced Recurrent Neural Network (SRNN) to extract the time series data's temporal characteristics of PMU data. Thirdly, based on the temporal characteristics, the Convolutional Neural Networks (CNN) and Attention mechanisms are used to extract the spatial features of these data. Finally, we sent the spatial features to the Fully Layer and used the softmax function to classify. Results: The proposed SRCAM in this paper has two advantages. One is that it implements the parallel computation on data by using the segmentation concept of SRNN, which reduces the computation time. The other is that using the Attention mechanism on CNN can make the spatial features more prominent. At the end of the paper, we do many comparative experiments between SRCAM and other models, such as some algorithms of Machine learning and soft computing. We take IEEE node data as experimental data and TensorFlow as an experimental platform. Experimental results show that the SRCAM model has an excellent performance of the detection of PDMA with high precision and accuracy. Conclusion: The superiority of SRCAM is theoretically and experimentally proved in this paper. As we expected, SRCAM showed great results in the application of PDMA detection.

Improved YOLOv4 Marine Target Detection Combined with CBAM

Marine target detection technology plays an important role in sea surface monitoring, sea area management, ship collision avoidance, and other fields. Traditional marine target detection algorithms cannot meet the requirements of accuracy and speed. This article uses the advantages of deep learning in big data feature learning to propose the YOLOv4 marine target detection method fused with a convolutional attention module. Marine target detection datasets were collected and produced and marine targets were divided into ten categories, including speedboat, warship, passenger ship, cargo ship, sailboat, tugboat, and kayak. Aiming at the problem of insufficient detection accuracy of YOLOv4’s self-built marine target dataset, a convolutional attention module is added to the YOLOv4 network to increase the weight of useful features while suppressing the weight of invalid features to improve detection accuracy. The experimental results show that the improved YOLOv4 has higher detection accuracy than the original YOLOv4, and has better detection results for small targets, multiple targets, and overlapping targets. The detection speed meets the real-time requirements, verifying the effectiveness of the improved algorithm.

Research on Intelligent Target Recognition Integrated With Knowledge

With the development of artificial intelligence technology, intelligent weapon systems that can automatically identify, lock on and strike targets have gradually appeared and can replace humans in executing simple decision-making commands. Target detection is a key part of intelligent weapons. At present, large-scale target detection has serious challenges such as long-tail data distributions, severe occlusion, and category ambiguity. The main detection algorithms only detect each independent area without considering the key semantic dependencies between objects. It has become a hot trend to apply deep learning to prior knowledge to form a model. This article uses both internal and external knowledge to instill a target detection system with human reasoning capabilities. Commonly used external embedded knowledge includes geometric relations, attributes, locations, etc. They have a common shortcoming in that they require large amounts of labeled data, and the integration costs are huge. The purpose of this article is to construct a general external prior knowledge module to guide network learning. By paying attention to the characteristics of each object in different semantic contexts, the characteristics of each object are adaptively enhanced, and the high-level semantics of all categories evolve on a global scale. Internal knowledge uses a convolutional attention module that can learn spatial and channel information at multiple scales. The experimental results show the superiority of our knowledge-YOLOv5. The proposed method achieved 1.7%, 2.2%, 1.1%, and 0.7% improvements over YOLOv5s, YOLOv5m, VOLOv51, and YOLOv5x, respectively, on the COCO data sets; and the proposed method also achieves a 0.9% improvement on the self-built data set. The trained lightweight model Knowledge-YOLOv5s is deployed on an NVIDIA Jetson TX2 through TensorRT acceleration, and the real-time detection frame is 20 ms, which meets the real-time detection requirements. This system can also be used as a module of an intelligent weapon system, which has certain referential significance for autonomous weapons and unmanned combat systems.

Structure Preserving Convolutional Attention for Image Captioning

In the task of image captioning, learning the attentive image regions is necessary to adaptively and precisely focus on the object semantics relevant to each decoded word. In this paper, we propose a convolutional attention module that can preserve the spatial structure of the image by performing the convolution operation directly on the 2D feature maps. The proposed attention mechanism contains two components: convolutional spatial attention and cross-channel attention, aiming to determine the intended regions to describe the image along the spatial and channel dimensions, respectively. Both of the two attentions are calculated at each decoding step. In order to preserve the spatial structure, instead of operating on the vector representation of each image grid, the two attention components are both computed directly on the entire feature maps with convolution operations. Experiments on two large-scale datasets (MSCOCO and Flickr30K) demonstrate the outstanding performance of our proposed method.