Search

The gangue recognition technology is one of the key technologies in the intelligent construction of mines. To address the problem of low accuracy of the gangue recognition model caused by low illumination and high dust environment at the working face and the difficulty of recognizing small target gangue, a coal gangue image recognition model based on CSPNet-YOLOv7 target detection algorithm is proposed. Cross Stage Partial Network (CSPNet) is used to improve the backbone feature extraction network of YOLOv7 model, optimize the gradient information to reduce the network parameters, while Recursive Feature Pyramid (RFP) and Switchable Auto Convolution (SAC) to replace the simple up and down sampling and normal convolution modules in the neck feature extraction network, and to enhance the generalization ability of the network by using three migration training for feature learning of different widths and depths. The experimental results show that the CSPNet-YOLOv7 model has an average accuracy mean of 97.53%, an accuracy rate of 92.24%, a recall rate of 97.91%, an F1 score of 0.95, a model parametric number of 30.85×106, a floating point operation count of 42.15×109, and a frame rate of 24.37 f/s transmitted per second, Compared to the YOLOv7 model, the average mean accuracy is improved by 7.46%, and the number of parameters and floating point operations are reduced by 17.23% and 60.41%, respectively, compared to the FasterRCNN-Resnet50, YOLOv3, YOLOv4, MobileNet V2 -YOLOv4, YOLOv4-VGG, YOLOv5s models. The CSPNet-YOLOv7 model has the highest average accuracy mean for coal gangue identification, while the number of parameters and floating point operations is small, which has a good balance between identification accuracy and speed. Finally, the CSPNet-YOLOv7 model is validated through downhole field tests, providing an effective technical means for accurate coal gangue identification.

Copyright of Coal Science & Technology (0253-2336) is the property of Coal Science & Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The distribution of forest-dominant tree species is crucial for ecosystem assessment. Remote sensing monitoring requires annual ground sample data, but consistent field surveys are challenging. This study addresses this by combining sample migration learning and machine learning for multi-year tree species classification in the Three Gorges Reservoir area in China. Using the continuous change detection and classification (CCDC) algorithm, sample data from 2023 were successfully migrated to 2018–2022, achieving high migration accuracy ( R 2 = 0.8303, RMSE = 4.64). Based on migrated samples, random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGB) algorithms classified forest tree species with overall accuracies above 70% and Kappa coefficients above 0.6. XGB. They outperformed other algorithms, with classification accuracy of over 80% and Kappa above 0.75 in almost all years. The final map indicates stable distribution from 2018 to 2023, with eucalyptus covering over 40% of the forest area, followed by horsetail pine, fir, cypress, and wetland pine. [ABSTRACT FROM AUTHOR]

Abstract As infrastructure-as-a-service clouds quickly grow, an increasing number of businesses and people are moving their application development to the cloud. The purpose of the research is to solve the problem of identifying memory anomalies in cloud virtual machines and improve the accuracy of the model in detecting abnormal situations. This paper presents a model for detecting virtual machine anomalies in IaaS cloud platform. The model considers the unique properties of monitoring metrics as time-series data and proposes an approach based on four important virtual machine monitoring metrics. The study also develops an adaptive anomaly detection system based on deep Q-network algorithms and migration learning principles for the variety of VM monitoring data in the cloud. The testing findings reveal that utilizing a Zoom layer with a 2-kernel size can increase detection accuracy to 96.7%. This demonstrates that a portion of the experimental data can extract the temporal features using the Zoom layer and different kernel sizes. The research model for anomaly detection had a classification accuracy of 99.8%. The deep Q-network model’s final anomaly detection accuracy varies from 96.7 to 98.6%. The outcomes of the research improved the system’s security and dependability, showed the worth of the overall framework design, and significantly decreased the number of resources needed for system operation and maintenance.

As infrastructure-as-a-service clouds quickly grow, an increasing number of businesses and people are moving their application development to the cloud. The purpose of the research is to solve the problem of identifying memory anomalies in cloud virtual machines and improve the accuracy of the model in detecting abnormal situations. This paper presents a model for detecting virtual machine anomalies in IaaS cloud platform. The model considers the unique properties of monitoring metrics as time-series data and proposes an approach based on four important virtual machine monitoring metrics. The study also develops an adaptive anomaly detection system based on deep Q-network algorithms and migration learning principles for the variety of VM monitoring data in the cloud. The testing findings reveal that utilizing a Zoom layer with a 2-kernel size can increase detection accuracy to 96.7%. This demonstrates that a portion of the experimental data can extract the temporal features using the Zoom layer and different kernel sizes. The research model for anomaly detection had a classification accuracy of 99.8%. The deep Q-network model's final anomaly detection accuracy varies from 96.7 to 98.6%. The outcomes of the research improved the system's security and dependability, showed the worth of the overall framework design, and significantly decreased the number of resources needed for system operation and maintenance. [ABSTRACT FROM AUTHOR]

In aquaculture, quantifying the real-time feeding behaviour of fish is essential for making feeding decisions. However, most existing methods for assessing fish appetite are inefficient and subjective. To address these issues, this study proposes an improved tilapia feeding level classification model with ResNet34. First, we introduce the attention module CBAM into the ResNet34 model to adjust the attention of the model according to the importance of different channel features and enhance the effective extraction of important features. We then used migration learning to transfer the knowledge learned from the source data (ImageNet dataset) to the tilapia ingestion image dataset, which allowed us to train the tilapia ingestion behaviour classification model faster while retaining the pre-trained model. Experimental results showed that the improved ResNet34 model in this study achieved an accuracy of 99.72%, an improvement of 7.84 percentage points over the original model. In addition, the model outperformed models such as MobileNetV2, AlexNet, VGG11, ShuffleNet_v2_x0_5 and ResNet18 in terms of accuracy, precision, recall and F1 scores.These results suggest that the proposed method can accurately identify feeding behavior of fish and provide a scientific basis for determining feeding amounts.

Current single-model methods for fine-grained image classification suffer from insufficient generalisation ability, while multi-model fusion methods suffer from weight curing. The study suggests and experimentally tests a dynamic weight multi-model fusion strategy for transfer learning-based fine-grained picture classification. The results of the experiment showed that the suggested fusion model enhanced recognition accuracy by 1.33%, 1.19%, and 0.83% compared to the single model on the medical dataset and 3.25%, 1.34%, and 7.28% on the agronomy dataset, respectively. Furthermore, when compared to the comparison method, the models under the proposed method of the study improved recognition accuracy by 0.18%, 0.61%, 0.43%, and 0.43% on the medical dataset, and the experimental time consumed was 3.25 minutes less than that of the sum-of-maximum-probabilities method; however, the fusion models of the proposed method of the study had higher recognition accuracy than that of the comparison met Overall, the proposed dynamic weight multi-model fusion method for fine-grained image classification using migration learning has better performance and generalisation ability, which can improve performance while reducing time cost, and has higher application value for the actual fine-grained image classification task.

The aim of this article is to propose a defect identification method for bare printed circuit boards (PCB) based on multi-feature fusion. This article establishes a description method for various features of grayscale, texture, and deep semantics of bare PCB images. First, the multi-scale directional projection feature, the multi-scale grey scale cooccurrence matrix feature, and the multi-scale gradient directional information entropy feature of PCB were extracted to build the shallow features of defect images. Then, based on migration learning, the feature extraction network of the pre-trained Visual Geometry Group16 (VGG-16) convolutional neural network model was used to extract the deep semantic feature of the bare PCB images. A multi-feature fusion method based on principal component analysis and Bayesian theory was established. The shallow image feature was then fused with the deep semantic feature, which improved the ability of feature vectors to characterize defects. Finally, the feature vectors were input as feature sequences to support vector machines for training, which completed the classification and recognition of bare PCB defects. Experimental results show that the algorithm integrating deep features and multi-scale shallow features had a high recognition rate for bare PCB defects, with an accuracy rate of over 99%. [ABSTRACT FROM AUTHOR]

The distribution of forest-dominant tree species is crucial for ecosystem assessment. Remote sensing monitoring requires annual ground sample data, but consistent field surveys are challenging. This study addresses this by combining sample migration learning and machine learning for multi-year tree species classification in the Three Gorges Reservoir area in China. Using the continuous change detection and classification (CCDC) algorithm, sample data from 2023 were successfully migrated to 2018–2022, achieving high migration accuracy (R2 = 0.8303, RMSE = 4.64). Based on migrated samples, random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGB) algorithms classified forest tree species with overall accuracies above 70% and Kappa coefficients above 0.6. XGB. They outperformed other algorithms, with classification accuracy of over 80% and Kappa above 0.75 in almost all years. The final map indicates stable distribution from 2018 to 2023, with eucalyptus covering over 40% of the forest area, followed by horsetail pine, fir, cypress, and wetland pine.

In this paper, the self-attention layer of a graph convolutional neural network is first constructed to output the important information in the network structure. The migration learning network model is established, and the sample data are preprocessed and trained sequentially. The final processing results are used as the initial data for abnormal power consumption detection. Introduce Bayes’ theorem to optimize the hyperparameters of the model. The optimized model is applied in the abnormal power consumption detection system to identify abnormal power consumption events and provide specific processing solutions. Through the detection of the system, it was found that the voltage of the test user dropped from a 100V cliff to about 20V in late November, which was determined by the system to be a power consumption abnormality, and, therefore, an operation and maintenance order was issued. The site survey revealed that the data was in line with the system detection. Calculating the power consumption information of another user, the phase voltage of this user stays around 85-100V, far below 150V, so the undercounting of power is verified for the user, and the amount of power that should be recovered is 201.22kW.

The current music recommendation is less efficient and cannot meet the needs of most people. This paper combines deep learning and migration learning to research music recommendations. It mainly utilizes the hybrid recommendation method of label recommendation and noise reduction autoencoder to achieve accurate music recommendations after extracting the features of music and the user’s preference features. In addition, in order to explore the recommendation effect of the model in this paper, the deviation between the measured score and the actual score of this model is compared and analyzed as well as the similarity difference of music with different features is explored in terms of HR, Recall, and NDCG index performance. The results show that the number of songs with a deviation of 2 or less in the predicted scores of this recommender system all account for 88% of the total number of songs, which is better than other models. In terms of HR, Recall, and NDCG, this paper’s model has better recommendation performance than other models, and the similarity of music recommendation between the same album and different albums is more than 0.65. This study is of great significance for eliminating the information barriers to enhance and explore the value of music data information.

In the continuous progress of mobile internet technology, medical image processing technology is also always being upgraded and improved. In this field, digital watermarking technology is significant and provides a strong guarantee for medical image information security. This paper offers a robustness zero watermarking strategy for medical pictures based on an Improved NasNet-Mobile convolutional neural network and the discrete cosine transform (DCT) to address the lack of robustness of existing medical image watermarking algorithms. First, the structure of the pre-training network NasNet-Mobile is adjusted by using a fully connected layer with 128 output and a regression layer instead of the original Softmax layer and classification layer, thus generating a regression network with 128 output, whereby the 128 features are extracted from the medical images using the NasNet-Mobile network with migration learning. Migration learning is then performed on the modified NasNet-Mobile network to obtain the trained network, which is then used to extract medical image features, and finally the extracted image features are subjected to DCT transform to extract low frequency data, and the perceptual hashing algorithm processes the extracted data to obtain a 32-bit binary feature vector. Before performing the watermark embedding, the watermark data is encrypted using the chaos mapping algorithm to increase data security. Next, the zero watermarking technique is used to allow the algorithm to embed and extract the watermark without changing the information contained in the medical image. The experimental findings demonstrate the algorithm's strong resistance to both conventional and geometric assaults. The algorithm offers some practical application value in the realm of medicine when compared to other approaches. [ABSTRACT FROM AUTHOR]

The goal of the work described here was to construct a deep learning–based intelligent diagnostic model for ophthalmic ultrasound images to provide auxiliary analysis for the intelligent clinical diagnosis of posterior ocular segment diseases. The InceptionV3–Xception fusion model was established by using two pre-trained network models—InceptionV3 and Xception—in series to achieve multilevel feature extraction and fusion, and a classifier more suitable for the multiclassification recognition task of ophthalmic ultrasound images was designed to classify 3402 ophthalmic ultrasound images. The accuracy, macro-average precision, macro-average sensitivity, macro-average F1 value, subject working feature curves and area under the curve were used as model evaluation metrics, and the credibility of the model was assessed by testing the decision basis of the model using a gradient-weighted class activation mapping method. The accuracy, precision, sensitivity and area under the subject working feature curve of the InceptionV3–Xception fusion model on the test set reached 0.9673, 0.9521, 0.9528 and 0.9988, respectively. The model decision basis was consistent with the clinical diagnosis basis of the ophthalmologist, which proves that the model has good reliability. The deep learning–based ophthalmic ultrasound image intelligent diagnosis model can accurately screen and identify five posterior ocular segment diseases, which is beneficial to the intelligent development of ophthalmic clinical diagnosis. [ABSTRACT FROM AUTHOR]

Drilling is one of the most classical machining operations. Real-time monitoring of drill wear can effectively testify whether the product fails to meet the specifications due to drill failure. This paper proposes a tool wear assessment and life prediction model based on image processing and deep learning methods, which works effectively for small sample datasets and for low-quality images. The normal areas and worn areas of the drill bits are extracted using the U-Net network and traditional image processing methods, respectively. Moreover, the original dataset is classified using the migration learning technique. The wear level of a drill bit can be accurately evaluated through experimental tests. Testing results show that the proposed method is more convenient and efficient than previous methods using manual measurements. These results can be applied to real-time drill wear monitoring, thus reducing part damage caused by tool wear. [ABSTRACT FROM AUTHOR]

In view that traditional manual feature extraction method cannot effectively extract the overall deep image information, a new method of scene classification based on deep learning feature fusion is proposed for remote sensing images. First, the Grey Level Co-occurrence Matrix (GLCM) and Local Binary Patterns (LBP) are used to extract the shallow information of texture features with relevant spatial characteristics and local texture features as well; second, the deep information of images is extracted by the AlexNet migration learning network, and a 256-dimensional fully connected layer is added as feature output while the last fully connected layer is removed; and the two features are adaptively integrated, then the remote sensing images are classified and identified by the Grid Search optimized Support Vector Machine (GS-SVM).The experimental results on 21 types of target data of the public dataset UC Merced and 7 types of target data of RSSCN7 produced average accuracy rates of 94.77% and 93.79%,respectively, showing that the proposed method can effectively improve the classification accuracy of remote sensing image scenes. [ABSTRACT FROM AUTHOR]

To solve the problem of robustness of encrypted medical image watermarking algorithms, a zero watermarking algorithm based on the discrete cosine transform (DCT) and an improved DarkNet53 convolutional neural network is proposed. The algorithm targets medical images in the encrypted domain. In this algorithm, DCT is performed on the encrypted medical image to extract 32-bit features as feature 1. DarkNet53, a pre-trained network, was chosen for migration learning for the network model. The network uses a fully connected layer and a regression layer instead of the original Softmax layer and classification layer, changing the original classification network into a regression network with an output of 128. With these transformations, 128-bit features can be extracted from encrypted medical images by this network, and then DCT is performed to extract 32-bit features as feature 2. The fusion of features 1 and 2 can effectively improve the robustness of the algorithm. The experimental results show that the algorithm can accurately distinguish different encrypted medical images and can effectively restore the original information from the encrypted watermarked information under traditional and geometric attacks. Compared with other algorithms, the proposed method demonstrates better robustness and invisibility. [ABSTRACT FROM AUTHOR]

In this paper, we construct models for convolutional neural networks and migration learning, conduct in-depth research on rice pest recognition methods, and design a plan based on convolutional neural networks and migration learning. The weight parameters obtained from the VGG16 model trained on the image dataset Image Net are migrated to recognize rice pests through the migration learning method. The convolutional and pooling layers of VGG16 are used as feature extraction layers. In contrast, the top layer is redesigned as a global average pooling layer and a SoftMax output layer, and some of the convolutional layers are frozen during training. The proposed method effectively improves the recognition accuracy of water to pest images and significantly reduces the number of model parameters. [ABSTRACT FROM AUTHOR]

To address the problem that the ant colony algorithm in solving Traveling Salesman Problem (TSP) has a slow convergence speed and easily falls into local optimum, an Ant Colony Algorithm with n- α -measure and Migration Learning is proposed. Firstly, the n- α -measure is proposed to map the least squares to the distance matrix of TSP nodes for improving the connection between nodes; second, multiple parameters in the ant colony algorithm are extended in multiple dimensions to improve the control of the algorithm over the pheromone; finally, a communication channel is established between multiple populations, and a transfer learning mechanism is proposed to exchange information. The simulation experiments of multiple cases in TSPLIB show that the improved algorithm balances diversity and the convergence speed, and effectively improves the quality of the solution. [ABSTRACT FROM AUTHOR]

The vital behaviors of Magang geese are closely related to their growth and welfare status. Therefore, it is very necessary to accurately identify the key vital behaviors of Magang geese in husbandry and production. This study aims to efficiently and accurately monitor the key behaviors of Magang geese in pens. An improved algorithm was also proposed for the vital behavior of Magang geese using Double Head-YoloX (referred to as MGBM-DH-YoloX). The detection efficiency of the network was enhanced to reduce the number of YoloX heads. A tradeoff was made to balance the foreground and background using the Focal loss objective function. Efficient network training was achieved using migration training. Other techniques were utilized to identify the vital behavior and count their key behaviors once at a fixed time in an efficient and accurate way, including the identification and rule analysis of key behaviors, such as feeding, drinking, rest and stress of Magang Geese. Firstly, MGBM-DH-YoloX enhanced the data of the Magang geese images with Mosaic and Mixup. And then the enhanced dataset was used to train the DH-YoloX (Double Head - YoloX) for the detection of the vital behavior of the waterfowl. Finally, the vital behavior of the Magang geese was counted every 25 frames. The experiment was conducted in a flock-reared environment with the Magang geese as the subject of the case study. A behavioral target detection test was conducted during this time. Waterfowl videos were collected on multiple days from 26 December 2021 to 04 January 2022, and from 10-12 May 2022 at the Nanwei Building Experimental Farm, School of Animal Science, South China Agricultural University, Tianhe District, Guangzhou City, Guangdong Province, China. A platform was also built to acquire the data for the behavior detection of geese. The color images of Magang geese were facilitated to integrate the environmental factors in the vicinity of the breeding pen (the height of the pen, as well as the width and height of the equipment box). The bi-directional difference frame was adopted to extract the key frames, in order to quickly extract the required picture production dataset from the video data. A total of 1 600 picture data was extracted from the video data, and selected for a total of 2 000 picture data. A total of 1 600 images were randomly used for the training and validation sets, whereas, 400 images were for the test set in the 10 days of continuous activity video. The results showed that the MGBM-DH-YoloX algorithm achieved an average accuracy of 98.98% mAP, a processing frame rate of 81 frame/s, and a memory consumption of 2 520.04 MB for the detection of Magang geese’s behaviors. Meanwhile, it was found that the geese were foraged less frequently, as they grow older, after monitoring the 10-day breeding data. The vital foraging and drinking behaviors simultaneously accounted for 83.74% of the total foraging time per day, indicating the overall companionship trend. There was also a decrease from 90.78% to 74.57%, indicating the gradual separation from the feeding and drinking behavior of Magang Geese with the increase of age. On the contrary, the resting behavior of the experimental geese increased slowly with the increase of age, indicating a gradual adaptation to the caging. The stress behavior was highly random in this experiment. There was an extremely serious emergency caused by irregular feeding, such as the random walking of personnel. Consequently, the MGBM-DH-YoloX can be expected as the video monitoring to intelligently extract the vital behavior of Magang geese. The finding can provide technical support for the automated monitoring of poultry in intelligent breeding supervision. [ABSTRACT FROM AUTHOR]

Orbital angular momentum (OAM) multiplexing technology is a highly promising approach that can significantly increase the data capacity of optical communication systems. However, traditional optical communication systems are prone to atmospheric disturbances such as turbulence, which can degrade transmission quality. To mitigate this issue, multimode fibers (MMFs) have been introduced to reduce the impact of turbulence on signals. Moreover, previous studies have primarily focused on the identification of single vortex beams, facing challenges in accurately recognizing multiplexed modes. To address this challenge, this chapter proposes a multi-label image classification optimization algorithm based on transfer learning. By utilizing the pre-trained MobileNet V2 model as a feature extractor, this network structure can accurately identify 8-bit, 16-bit, and 24-bit multiplexed OAM from speckle patterns in multimode fibers, even with small sample datasets, achieving classification accuracies exceeding 95%. This method overcomes the limitations of traditional optical communication systems that are susceptible to atmospheric disturbances, providing new possibilities for long-distance transmission and increased data capacity in optical communication systems. • We use MMF to transmit vortex light, which significantly reduces the impact of atmospheric turbulence. • A multi-label image classification optimization algorithm based on transfer learning, ML-OAM-DemuxNet, is proposed to identify speckle patterns. • Cutting the center part of the speckle pattern for recognition can greatly reduce the computing cost of the network while maintaining the accuracy [ABSTRACT FROM AUTHOR]

The joint torque estimation model based on surface electromyography (sEMG) signals needs to recalibrate the model parameters when used for joint torque estimation at different volunteers. However, the feature transferability drops significantly in higher task-specific layers, which reduce the performance of the model. The generalization of joint torque estimation models among different volunteers has become a challenge. In this paper, a fast calibration method based on a deep adaptive regression network is proposed to reduce the frequency and time of model recalibration when volunteers change. The proposed method associates source domain (SD) data with target domain (TD) data through migration learning. The maximum mean difference (MMD) adaptive method is used to migrate the model trained in the SD dataset to the TD dataset, which avoids the recalibration required in the joint torque estimation model. The experimental results show that the proposed method can improve the migration ability of the model used for estimating joint torque and achieve the generalization application of the model on different volunteer datasets. Moreover, the proposed method can reduce the calibration operations and iterations of the model caused by volunteer changes and improve the speed of joint torque estimation. The optimal normalized root mean square error (NRMSE) and Pearson correlation coefficient (ρ) of knee extension/flexion joint torque estimation reached 0.02198/0.02565 and 0.9892/0.9839, respectively. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Cotton Textile Technology is the property of Cotton Textile Technology Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Coal Science & Technology (0253-2336) is the property of Coal Science & Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The current most advanced CNN-based detection models are nearly not deployable on mobile devices with limited arithmetic power due to problems such as too many redundant parameters and excessive arithmetic power required, and knowledge distillation as a potentially practical model compression approach can alleviate this limitation. In the past, feature-based knowledge distillation algorithms focused more on transferring the local features customized by people and reduced the full grasp of global information in images. To address the shortcomings of traditional feature distillation algorithms, we first improve GAMAttention to learn the global feature representation in images, and the improved attention mechanism can minimize the information loss caused by processing features. Secondly, feature shifting no longer defines manually which features should be shifted, a more interpretable approach is proposed where the student network learns to emulate the high-response feature regions predicted by the teacher network, which increases the end-to-end properties of the model, and feature shifting allows the student network to simulate the teacher network in generating semantically strong feature maps to improve the detection performance of the small model. To avoid learning too many noisy features when learning background features, these two parts of feature distillation are assigned different weights. Finally, logical distillation is performed on the prediction heads of the student and teacher networks. In this experiment, we chose Yolov5 as the base network structure for teacher–student pairs. We improved Yolov5s through attention and knowledge distillation, ultimately achieving a 1.3% performance gain on VOC and a 1.8% performance gain on KITTI. • We use knowledge distillation and attention for yolov5s to improve performance. • We propose the GAMF module that captures both global and local features. • We propose the RFT and join SSIM to shift response features and background features. [ABSTRACT FROM AUTHOR]

Aiming at the problems of low accuracy, slow speed and low defect identification of industrial site fabric defect detection, a improved fabric defect detection algorithm based on ResNet50 network is proposed. Firstly, the data set was preprocessed, and the model training set was enhanced by cutting the data samples, including the fault-free and 8 common defect categories. Then, the structure of Resnet50 network was improved to extract the weight parameters pre-trained on the large data set ImageNet for transfer learning. Finally, the defect detection and recognition model obtained by hyper parameter training was adjusted repeatedly. The results of several groups of comparative experiments show that the recognition accuracy of the improved model for normal fabric and 8 common defects reaches 96. 32%, which is 4.2% higher than that of the standard model, and the speed is doubled. The test in different fabric defect data sets has the best comprehensive performance, strong generalization ability and good robustness, which can meet the requirements of fabric defect detection in industrial production sites. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Guangxi Normal University - Natural Science Edition is the property of Gai Kan Bian Wei Hui and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In order to solve the problem of network intrusion detection, improve detection accuracy and reduce false positive rate, this paper proposed a network intrusion detection method based on deep transfer learning. This method used unsupervised learning deep self-encoder for transfer learning to realize network intrusion detection. Firstly, it modeled the deep transfer learning problem, and then modeled the deep transfer learning problem. The transfer learning framework implemented encoding and decoding by embedding layer and label layer, and shared the weight of encoding and decoding by source domain and target domain for knowledge transferring. In the embedding layer, it compelled the distribution of source domain and target domain data to be similar by minimizing the KL divergence of embedded instances between domains. In the label coding layer,it coded and classified the label information of source domain by using the softwaremax regression model. The experimental results show that this method can implement network intrusion detection, and its performance is better than other intrusion detection methods. [ABSTRACT FROM AUTHOR]

How to learn and identify new categories from a small number of training samples is a challenging problem for deep neural networks. For how to solve the problem of few-shot learning, this paper comprehensively summarized the existing few-shot learning methods based on deep neural networks, which covered various aspects such as models used in the methods, datasets and evaluation results. Specifically, for the few-shot learning method based on deep neural network, this paper divided it into four categories, named data enhancement method, migration learning method, metric learning method and meta-learning method. For each category, this paper further divided it into sub-categories and conducted a series of comparisons between each category and method to show the pros and cons of the various methods and their respective characteristics. Finally, the paper highlighted the limitations of existing methods and pointed to future research directions in the field of few-shot learning. [ABSTRACT FROM AUTHOR]

In today's rapid development of science and technology, the manual work in the factory has been gradually replaced by machines, and the process of industrial intelligence has been further deepened. Workpiece recognition is the use of machine learning, computer vision and other technologies to identify the target workpiece, and the robot intelligent capture is a higher level of operation of the workpiece recognition, which is the key to realize the intelligentization of industrial robots. Due to the complex environmental factors and the diversity of the shape and size of the objects to be grasped, the accuracy and efficiency of the workpiece recognition are not ideal at this stage, and intelligent crawling is even more difficult to talk about. Aiming at the above problems, this paper builds a crawling planning model based on the convolutional neural network and the grasping pose mapping rules. Based on the established crawling planning model, the sampling candidate grabbing algorithm is designed and the migration learning method is adopted. The pretrained convolutional neural network for image recognition on the ImageNet dataset was migrated to the capture detection task of the Carnegie Mellon dataset. Experiments show that the network model proposed in this paper performs well, and its correct crawl rate is as high as 81.27%. This is to achieve a more stable and reliable identification and intelligent crawling work. [ABSTRACT FROM AUTHOR]

Traditional detection methods are not sensitive to small-sized tomato organs (flowers and fruits), because the immature green tomatoes are highly similar to the background color. The overlap among fruits and the occlusion of stems and leaves on tomato organs can lead to false and missing detection, which decreases the accuracy and generalization ability of the model. Therefore, a tomato organ recognition method based on improved Feature Pyramid Network was proposed in this paper. To begin with, multi-scale feature fusion was used to fuse the detailed bottom features and high-level semantic features to detect small-sized tomato organs to improve recognition rate. And then repulsion loss was used to take place of the original smooth L1 loss function. Besides, Soft-NMS (Soft non-maximum suppression) was adopted to replace non-maximum suppression to screen the bounding boxes of tomato organs to construct a recognition model of tomato key organ. Finally, the network was trained and verified on the collected image data set. The results showed that compared with the traditional Faster R-CNN model, the performance was greatly improved (mean average precision was improved from 90.7 to 99.5%). Subsequently, the training model can be compressed so that it can be embedded into the microcontroller to develop further precise pesticide targeting application system of tomato organs and the automatic picking device. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Recently, deep learning has introduced new prospects to numerous practical applications such as image recognition, robot navigation, gene engineering, language processing, and traffic sign identification. Several network models including AlexNet, VGGNet, GoogLenet and ResNet, have achieved milestone contributions while relying on massive computing resources. However, when faced with a small number of labeled examples, especially in the case of unbalanced datasets, the cumulative error and time-consuming convergence reduce their efficacy. Inspired by the convolutional output layer with a 1 × 1 kernel, a convolutional nonlinear transfer approach with partial cooperating (CNN-COL) is proposed to address this challenge. Meanwhile, a novel method for data augmented balance can enhance the influence of small and unbalanced samples in the CNN-COL. Related experiments show that the proposed CNN-COL can effectively improve the quality of a dataset and achieve superior performance with respect to traffic sign identification based on a small and type-unbalanced dataset. [ABSTRACT FROM AUTHOR]

In order to solve the problem in extracting features manually in the application of remote sensing, which is time-consuming, labor-intensive and low accuracy, and is hard to acquire the annotated data, the tree species classification algorithm of convolutional neural network remote sensing image based on migration learning is proposed. Firstly, the lneven-v3 model trained on ImageNet is introduced to extract the image features of tree species. A new fully connected layer and Softmax layer are trained to change the figure of nodes in the output layer through the AID standard dataset and the self-collecting dataset of experimental forest farm of Northeast Forestry University. Secondly, the phenomenon of overfitting is improved through the introduction of the Dropout function. Finally, the model is optimized by back propagation. The experiment results show that the overall classification accuracies on two data sets are 98. 8% and 97. 2%, respectively, and the Kappa coefficients are 0. 987 and 0. 984, respectively. The experiment results indicate that the algorithm mentioned above can not only reduce the number of parameters of the convolutional neural network, but also improve the accuracy of target classification and recognition. Compared with the traditional methods of tree species classification, this algorithm can automatically extract the characteristics of tree species, and the problem that the high accuracy depends on the number of experimental samples is solved, and this algorithm has high application value in tree species classification. [ABSTRACT FROM AUTHOR]

Accurate detection of flames is an important prerequisite for the implementation of many security measures. To address the problems of low detection accuracy and difficulty in detecting small target flame in the current mainstream flame detection algorithms, we propose a small target flame detection algorithm incorporating an attention mechanism. First, we optimize the self-acquired dataset with data enhancement and structural similarity algorithms to improve the quality of the dataset and enhance the training efficiency of the model; then, a coordinate attention mechanism is introduced in the feature extraction stage of the original you-only-look-once (YOLOv7) network to improve the ability of the model to extract useful feature information and help the model to more accurately; and then, a small target flame detection layer is introduced at the output of the YOLOv7 network to improve the model's ability to detect small flames. Finally, the parameters of each layer of the improved model are initialized through model pre-training and migration learning to reduce the risk of model overfitting and accelerate the convergence speed of model training. The experimental results show that the average detection accuracy of the flame detection algorithm designed in this paper reaches up to 89.5%, which is 8.7% better than the detection accuracy of the original YOLOv7 algorithm and is innovative and practical. [ABSTRACT FROM AUTHOR]

In the current natural environment, due to the complexity of the background and the high similarity of the color between immature green tomatoes and the plant, the occlusion of the key organs (flower and fruit) by the leaves and stems will lead to low recognition rates and poor generalizations of the detection model. Therefore, an improved tomato organ detection method based on convolutional neural network (CNN) has been proposed in this paper. Based on the original Faster R-CNN algorithm, Resnet-50 with residual blocks was used to replace the traditional vgg16 feature extraction network, and a K-means clustering method was used to adjust more appropriate anchor sizes than manual setting, to improve detection accuracy. The test results showed that the mean average precision (mAP) was significantly improved compared with the traditional Faster R-CNN model. The training model can be transplanted to the embedded system, which lays a theoretical foundation for the development of a precise targeting pesticide application system and an automatic picking device.

As single source domain lo a single target domain and the category number must be consistent constrained the traditional migration learning clustering algorithms,effect of clustering was not good,this paper proposed a clustering algorithm of learning across the source domain. This algorithm had three advantages: a)The new algorithm not only expanded the number of the source domain but also canceled the limits of category number of the source domain, it could adaptively select tbe source domain to study, so the migration learning ability of tbe algorithm can get larger advance; b) As tbe source domain knowledge used by tbe algorithm would not expose the original data, so the algorithm had good data privacy protection of source domain; c) Adjusting the balance parameters could make the algorithm degrade into the traditional clustering algorithm, so clustering performance of the algorithm was guaranteed. Through the experiments on simulated data sets and real data sets, this paper verifies that the proposed algorithm has better migration ability compared with the traditional migration learning clustering algorithm . and the clustering performance and robustness have larger advance. [ABSTRACT FROM AUTHOR]

The popularization of electronic clinical medical records makes it possible to use automated methods to extract high-value information from medical records quickly. As essential medical information, oncology medical events are composed of attributes that describe malignant tumors. In recent years, oncology medicine event extraction has become a research hotspot in academia. Many academic conferences publish it as an evaluation task and provide a series of high-quality annotation data. This article aims at the characteristics of discrete attributes of tumor-related medical events and proposes a medical event. The standard extraction method realizes the combined extraction of the primary tumor site and primary tumor size characteristics, as well as the extraction of tumor metastasis sites. In addition, given the problems of the small number and types of annotation texts for tumor-related medical events, a key-based approach is proposed. A pseudo-data-generation algorithm that randomly replaces information in the whole domain improves the transfer learning ability of the standard extraction method for different types of tumor-related medical event extractions. The proposed method won third place in the clinical medical event extraction and evaluation task of the CCKS2020 electronic medical record. A large number of experiments on the CCKS2020 dataset verify the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

• A hybrid SIFC method is proposed to determine the unknown state or new load. • The isolated forest method is introduced to improve the judgment of unknown loads. • A shared parameter migration method is proposed to realize model self-updating. With the development of smart grid technology, optimization of energy structure, improvement of power efficiency and reduction of power consumption have become the major development trends. As a key process of power structure analysis, load identification is gaining increasing attention in smart grids. Although lots of load identification methods have been proposed, the introduction of unknown states or new load types is still a challenge for precise identification. In this paper, a similarity calculation method of the space convex hull overlap rate is proposed to deal with the identification of unknown loads based on the low-dimensional feature space model of Siamese neural networks. Moreover, transfer learning is introduced to realize the category-added learning of load pre-identification model for unknown load. The ultimate aim was to establish the identification and rapid modeling of unknown loads. The public datasets Plaid is used to verify the performance of the proposed method. As a result, the high accuracy of load identification for unknown load is proved. [ABSTRACT FROM AUTHOR]

In the current natural environment, due to the complexity of the background and the high similarity of the color between immature green tomatoes and the plant, the occlusion of the key organs (flower and fruit) by the leaves and stems will lead to low recognition rates and poor generalizations of the detection model. Therefore, an improved tomato organ detection method based on convolutional neural network (CNN) has been proposed in this paper. Based on the original Faster R-CNN algorithm, Resnet-50 with residual blocks was used to replace the traditional vgg16 feature extraction network, and a K-means clustering method was used to adjust more appropriate anchor sizes than manual setting, to improve detection accuracy. The test results showed that the mean average precision (mAP) was significantly improved compared with the traditional Faster R-CNN model. The training model can be transplanted to the embedded system, which lays a theoretical foundation for the development of a precise targeting pesticide application system and an automatic picking device. [ABSTRACT FROM AUTHOR]