Search

In view of the significance of deep learning, this paper reviewed the research and development on deep learning. Firstly, this paper summarized the advantage of deep learning, and illustrated the necessity of introducing deep learning. Secondly, it described three kinds of typical deep learning models, included convolutional neural network model, deep belief network model, and stacked auto-encoder network model. Thirdly, it reviewed new research and development on deep learning in recent years, included the choice of initialization methods, the number of network layers, and activation function, model structure, learning algorithms, and practical application. Finally, it presented the problems to be solved in aspects of theoretical analysis Representation and model of data, feature extraction, training and optimization, and research extension. [ABSTRACT FROM AUTHOR]

In order to predict the traffic congestion in city, this paper proposed the prediction model based on the deep learning. The traffic data like flow parameters, environmental situation, times of day and so on were summarized and composed into the feature vector, and defined four kinds of traffic flow pattern. It learned the unlabelled traffic dataset by the auto-encoder network to build the hidden layer function and generate the new feature dataset, and used the softmax regression to classify the traffic pattern based on the labeled feature dataset learning. Demonstrated by the simulation, the prediction model has better performance than the model without the feature learning, the average prediction accuracy can reach 85%. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Image & Graphics is the property of Editorial Office of Journal of Image & Graphics 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.)

Visual sentiment analysis intends to describe images and videos in a subjective and emotional way and to detect and express the information with computers. With the emergence of social media and the rise of plenty of training data, technological revolution like deep learning has brought new opportunities and challenges into this field. This paper reviewed the key technologies including visual feature extraction, emotional space construction and mapping from visual features to emotion in traditional visual sentiment analysis on the basis of the framework for visual emotional semantic extraction. Then it discussed visual sentiment analysis based on middle representation and object detection and the application of deep learning in visual sentiment analysis under the background of big data. Finally it summarized the existing problems of visual sentiment analysis and gave inspirational future research directions. [ABSTRACT FROM AUTHOR]

Data fusion algorithms widely used BP neural network to extract and classify the node data features in wireless sensor networks. In order to overcome the shortcomings of BP neural network leading to poor performance for data fusion,such as low convergence speed,local optimal and bad generalization ability, this paper proposed a data fusion algorithm SAESMDA combined with deep learning technology and wireless sensor network clustering routing protocol. SAESMDA used deep learning model SAESM based on stacked autoencoder (SAE) instead of the BP neural network, algorithm firstly trained SAESM in sink node and generated clusters for network, then used SAESM to exacted node data features in cluster nodes, finally the data features in the same class would be fused and sent to sink node by cluster heads. Simulation experiments show that compared with BPNDA based on the BP neural network, SAESMDA has a higher feature extraction and classification accuracy with the similar network energy consumption. [ABSTRACT FROM AUTHOR]

In view of the problem that the training sample size is large and the complex function fitting ability is weak in shallow classification method, this paper proposed a pedestrian detection method based on improved deep belief network classification algorithm. Firstly, RBM with T distribution function show layer nodes built an improve way of input, which could change information of pedestrian feature to Bernoulli distribution and recognized Bernoulli distribution; in addition, set up the middle layer RBM structure, which achieved data transfer between the hidden layer structure, and kept the key information. Finally, this paper used the BP neural network for output of classifier, which could back propagation the information of misclassification, and minor adjustments parameters of classification structure. Experimental results show that the improved deep belief network pedestrian detection method is better than other shallow classic algorithms, and the real time also can meet the needs of practical use. [ABSTRACT FROM AUTHOR]

Copyright of Geomatics & Information Science of Wuhan University is the property of Geomatics & Information Science of Wuhan University 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.)

With the improvement of the Internet, face recognition has been widely adopted in the security area. Traditional face recognition methods have slowly development in the security area due to their poor generalization ability and inefficiency of dealing with the particularly complex function. In order to improve the accuracy of face recognition, this paper presents a new hybrid classifier model by combining deep Boltzmann machine (DBM) with discriminative restrict Boltzmann machine (DRBM) and compares with these traditional methods. This paper designs a sound experiment on the CMU-PIE face dataset to demonstrate the efficiency of the hybrid model. The results show that the hybrid classifier model has better recognition effect compared to the traditional methods on the two selected datasets, which would own more social benefits and help management decisions. [ABSTRACT FROM AUTHOR]

This paper tried to propose a method to solve the problem of sentiment classification by integrating word embedding and convolutional neural network (CNN). First of all, the method accomplished a training process with skip-gram model to generate word embedding of each word in the dataset. Then, it created a two-dimensional feature matrix which was the combination of word embedding of each word in a training sample as the input of CNN model. Each iteration process of training, entries of feature matrix would also update as part of model parameters. Secondly, this paper proposed a CNN structure which was mainly composed of three different sizes of convolution kernels so as to complete the automatic extraction process of a variety of local abstract features. Compared with traditional machine learning algorithms, the proposed word embedding and CNN based sentiment classification model has successfully improved classification accuracy by 5.04%. [ABSTRACT FROM AUTHOR]

This paper reviewed the deep learning algorithms and their applications. It elaborated the greedy layer training algorithm which used the fine-grained back-propagation (BP) learning following the layer-wise pre-training on each restricted Boltzmann machine (RBM) layer. After comparing and analyzing the three ways of gradient descent in the BP algorithm, this paper suggested applying stochastic gradient descent in online learning and adopting stochastic mini-batch gradient descent in static offline learning. It summarized the characteristic of the net-work structure in deep learning and recommend the design of state-of-art five-layer net-work architecture. It also analyzed the necessity of the nonlinear activation function in feedforward neural net-works and the advantages of the common activation functions, and recommended using ReLU activate function. Finally, the paper provided a brief summary of features and application scenarios of emerging deep neural net-works such as deep CNN( convolutional neural net-works), deep RNNs( recurrent neural net-works) and LSTM(long short-termmemory net-works), as well as the potential directions of future deep learning applications and research. [ABSTRACT FROM AUTHOR]

Due to the most health problems of the building structure damage is cumulative,which is difficult to be detected,the complexity of the actual structure and the ambient noise makes it more difficult to do the structural health monitoring,while the existing method requires a lot of data in the training model,but in practice for the tag data it is very complex. To overcome this problem,by the wireless sensor networks,this paper used the sparse coding to achieve bridge structural health monitoring,and through a large number of unlabeled examples of feature extraction in achieving sparse coding algorithms based on training, it realized the data dimensionality reduction and unlabeled data preprocessing. Secondly, it used the deep learning algorithm to predict the bridge structural health monitoring category,and also used the linear conjugate gradient-based optimization algorithm to improve the Hessian optimization, and used the semi definite Gauss-Newton Hessian matrix to replace uncertain Hessian matrix,which used the secondary target combinations to achieve deep learning algorithm efficiency. Experimental results show that the depth of the structural safety of the bridge learning detection algorithms achieves a high-precision structural health monitoring of ambient noise levels under sparse coding mentioned. [ABSTRACT FROM AUTHOR]

Considering the problems of the traditional facial expression recognition algorithm with poor robustness and easily disturbed by the face identity information, this paper proposed a face recognition algorithm based on denoising AutoEncoders. First, the method detected face area from per picture and made sure all of the face areas were the same scale. Then it constructed the model based on stacked denoising AutoEncoders for pre-training data. Finally, in order to avoid the overfitting problem that was caused by the lack of training samples,it adopted the Dropout technology in the last full connection layer of the deep network model. The experimental results on CK + database, JAFFE database and Yale database achieved the high accuracy rate. So that, this method has stronger robustness and capacity of resisting disturbance of the face identity information. [ABSTRACT FROM AUTHOR]

Partial least squares has been widely used in the multiple variable analysis. However, partial least squares method using principal component analysis, cannot express the nonlinear characteristic. and accuracy is low in the nonlinear data. Therefore, this paper proposed an analysis and predicting method of deep learning combining with PLS. The method could extract nonlinear structure of feature space by sparse autoencoder of deep learning and replace the components in PLS with the extracted components,forming a model which could adapt to nonlinear dose-effect relationship. It analyzed and processed the data of large chengqi decoction.ma xing shi gan tang.puerariae and Scutellariae and coptidis decoction,UCI data set. The experimental results show that,the deep learning and PLS method can well reflect the characteristics of TCM data. [ABSTRACT FROM AUTHOR]

In view of the existing prediction methods fail to fully reveal the nature of the traffic flow, this paper proposed a short-term traffic flow prediction model based on deep learning. The method combined the deep belief network( DBN) model and support vector regressionf SVR) classifier as predictive model. It removed the trend of the traffic flow by using data difference, extracted traffic flow features by deep belief network model, carried and the traffic flow prediction out with support vector regression in the top level of the network. Experiment results with actual traffic flow data show that the proposed method has a higher accuracy compared with others, prediction performance increases by 18. 01 % .and it is an effective traffic flow' prediction method. [ABSTRACT FROM AUTHOR]

With the rapidly development of natural language processing science, sentiment analysis as one of its important branches are widely used in social network platform. Especially the micro-blogging which are wide dissemination and contains a wealth of information on the emotional and highly favored by scholars. Due to the micro-blogging widely spread and contains rich human emotional information, it quickly become the research object of the Chinese and foreign scholars. To analysis the expression of human emotion in the micro-blogging and even digging its inherent sentiment, this paper made a further on the de-noising auto-encoder( DAE) model and explored a new method to improve it. The characteristics of DAE was to achieve the reduction of the original input under the interference of noise,while the advantage of the new model were take into account the diversity and complexity of noise,and to strengthen the resilience of the original features of the model through the deep learning training,then the result was overcame the original input noise which could not prediction. At last,by separately using SVM, DAE model and the improved model made sentiment analysis experiment, comparing the classification results indicate that the improved model(IDAE) are more accurate in micro-blogging sentiment analysis. Moreover, its anti-interference ability and robustness has improved. [ABSTRACT FROM AUTHOR]

In order to reduce the loss of dimension reduction in high dimensional image dataset and improve the effect of data visualization, this paper proposed a new image dataset visualization method based on the combination of deep features and nonlinear dimension reduction. This method firstly designed and trained a convolutional neural network model,and the single model on the MNIST dataset achieved the highest recognition accuracy. Secondly,it used the high precision model to extract the deep middle layer features as an effective representation of the image dataset. Finally,it used nonlinear dimension reduction method to reduce the features into two-dimensional and realized data visualization. Experimental results show that this method can effectively avoid the loss of traditional dimension reduction in image dataset visualization, and the visualization result is very obvious. [ABSTRACT FROM AUTHOR]

In ihe task of relation extraction, traditional statistical methods have difficulties in getting perfect features manually, while deep learning methods strongly depend on the representational capacity of single word vector. To solve the problems mentioned above, this paper proposed a novel model called multi-channel convolutional neural network (CNN). Firstly, the model used different word vectors to represent sentence as input for different channels. Secondly, it extracted features automatically through CNN. Finally, it obtained relation types by using softmax classifier. The proposed model could capture more semantic information thus it learned more distinctive features automatically comparing to other models. The experiment results on Sem E-val-2010 Task 8 datasets show that the proposed model is more suitable for relation extraction task than normal models which only utilize single word vector. [ABSTRACT FROM AUTHOR]

In view of the traditional decision tree need to predefine the features before classifying data, and in order to solve the bottleneck problems of processing speed and resource allocation when dealing with massive astronomical data, considering the strong representation learning of deep learning and good performance of processing huge amounts of data on Spark, this paper proposed a parallel deep neural decision tree based on Spark. And then it was applied on the astronomy star/galaxy separation problem. The results show that, this algorithm can scale well with cluster size as it can dramatically decrease the training time of model and enhance the ability of processing massive astronomy data with it. Moreover, it obtains a better classification accuracy on the star/galaxy separation problem by addressing the decision tree to learn the proper representations of input data and the final classifiers in a joint manner. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Signal Processing is the property of Journal of Signal Processing 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 Signal Processing is the property of Journal of Signal Processing 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.)

Tea is one of important cash crops in China. Computer vision plays an important role in pest detection. Automatic classification of insect species in field is more difficult than the generic object classification because of complex background in filed and high appearance similarity among insect species. In this paper, we proposed an insect recognition system on the basis of image saliency analysis and a deep learning model, i.e. convolutional neural network (CNN), which has a good robustness with avoiding the features selected by artificial means. In image saliency analysis, we segmented the original images into super-pixel regions firstly. Then we quantized each RGB (red, green, blue) color channel and made them have 10 different values, which reduced the number of colors to 1000, and sped up the process of the color contrast of the pest objects and the background at region level. Finally, we obtained the saliency value of each region by combining their color contrast and spatial distances. The saliency values of all regions in each image were used to construct a saliency map, which was offered as the initial area for GrabCut algorithm to define the segmentation result and localize the pest object. The images after localization were quantized to 256×256 dpi for CNN training and classifying. CNN was trained end to end, from raw pixels to ultimate categories, thereby alleviating the requirement to manually design a suitable feature extractor. Based on theoretical analysis and experimental evaluation, we optimized the critical structure parameters and training strategy of CNN to seek the best configuration. The overall architecture included a number of sensitive parameters and optimization strategies that could be changed. We determined the local receptive field size, number, and convolutional stride as 7×7 dpi, 64 and 4, respectively. Dropout ratio for the fully-connected layers was 0.7. The loss function Softmax was fit for the pest classification system. To further improve the practical utility of CNN, we focused on structural changes of the overall architecture that enabled a faster running with small effects on the performance. We analyzed the performance and the corresponding runtime of our model by reducing its depth (number of layers) and width (number of convolution kernel in each layer). Removing the fully-connected layers (FC6, FC7) made only a slight difference to the overall architecture. These layers contained almost 90% of the parameters and when they were removed, the memory consumption decreased to 29.8 MB. But, removing the intermediate convolutional layers (Conv2, Conv3, Conv4, Conv5) resulted in a dramatic decrease in both accuracy and runtime. This suggested that the intermediate convolutional layers (Conv2, Conv3, Conv4, Conv5) constituted the main part of the computational resource, and their depth was important for achieving good results. We then investigated the effects of adjusting the sizes of all convolutional layers, and the filters in each convolutional layer were reduced to 64 each time. Surprisingly, all architectures showed significant decreases in running time with relatively small effects on performance. Finally, we determined the convolution kernel numbers of Conv2-5: 64-192-192-64. On the test set of tea field images, the architecture before and after shrinking respectively achieved the average accuracy (AA) of 0.915 and 0.881, respectively, superior to previous methods for pest image recognition. Further, after optimization the running time reduced to 0.7 ms and the memory required was 6 MB. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Signal Processing is the property of Journal of Signal Processing 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.)

User complaints prediction model could effectively lower complaints rate of telecom users which was crucial for enterprise to improve customer satisfaction and competitiveness. During the process of model training，due to the defects of artificial design features and the unpredictable complexity in the design process，the accuracy of model prediction and the efficiency of the design features couldn't have a lot of ascension. For the above problems，this paper proposed a user complaints prediction model based on deep learning. By deep network characteristics study session，the model could automatically study the senior nonlinear combination features from the original telecommunication users' data which was suitable for classifier classification and import high-level features into traditional classifier to improve the precision of the model. By analyzing the experimental results，the prediction model' s AUC index has increased by 7. 1% compare to previous user complaints model. It has demonstrated that the effectiveness of the model automatic study features and availability of deep learning in telecom big data domain. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Image & Graphics is the property of Editorial Office of Journal of Image & Graphics 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 Image & Graphics is the property of Editorial Office of Journal of Image & Graphics 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 Remote Sensing is the property of Editorial Office of Journal of Remote Sensing & Science Publishing Co. 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 Signal Processing is the property of Journal of Signal Processing 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 Telecommunication Engineering is the property of Telecommunication Engineering 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 the era of big data, many of traditional machine learning methods of disposing unlabeled raw voice data have become less applicable. At the same time, deep learning models can directly process unlabeled data because of its powerful capability of modeling to deal with the massive data, and has become a hot research in the field of speech recognition. To begin with, this paper analyzed and summarized the state-of-the-art deep learning of models. And then,it discussed the applications to speech recognition with speech features extraction and acoustic modeling. Finally,it concluded the problems faced and development orientation. [ABSTRACT FROM AUTHOR]

In view of solving the problems of poor performance in the traditional collaborative filtering recommendation algorithm in dealing with the cold start and sparse data, this paper proposed a method that combined the deep model of the stacked denoising autoencoders with the nearest neighbor recommendation method. Aiming to form new hybrid recommendation model, treating the autoencoder as the basic unit, training process transformed a unsupervised learning problem. By combining sparse coding algorithm, the denoising criteria, extreme learning machine and using the idea of layer by layer autoencoder. The abstract features extracted from the final model applied to the nearest neighbor algorithm to predict the scoring. The experimental data sets of various models show that when the sparsity is less than 8% , compared with the cosine similarity model and Pearson similarity model, hybrid recommendation model experimental results are increased by 11.3% and 21. 1% , and compared with the potential matrix decomposition model , number of iterations for convergence that mixed model required is less nearly 30%. In the three groups of comparative experiments,the stability of the hybrid model also is the best, thus validating faster convergence speed of mixed model and impressive effect in deal with data sparsity and cold start. [ABSTRACT FROM AUTHOR]

This paper proposed a new method called stacked denoising autoencoder based on collaborative filtering after studying the way of improving precision of recommended algorithm. SDAE was a mul-layer deep learning model which could learn significant dependencies from input information. It used ratings given to items by users to learn the latent feature,got results of dimensionality reduction after PCA to calculate attribute similarities. Then it added the grade similarities computed with concealed feature as the final result. Finally, it emploied similarities between items to generate the top-N recommendation list. Experiments on MovieLens show that the proposed algorithm has higher recall rate and can settle the missing ratings and cold start problems in some way. [ABSTRACT FROM AUTHOR]

Recognition of the main organs of tomato from plant images has important significance for the detection of plant diseases and insect pests, the application of targeted pesticides, and the development of intelligent agricultural machinery. The traditional object recognition methods which are limited by hand-engineered features extraction process and the selection of filter operator types, have poor generality and low recognition accuracy. The convolutional layers of convolutional neural network (CNN) can achieve automatic extraction of image features. In view of this, we propose a tomato main organs recognition method which relies on deep convolutional neural network (DCNN) in this paper. Under different illumination conditions, and considering the morphological diversity of tomato organs, 5 kinds of images including flower, fruit, stem, leaf, and environment background and the plant images are collected, which are used as training and testing samples. Under the premise of maintaining the basic structure of VGGNet, by adjusting the number of convolutional kernels and the depth of convolutional layers, 10 kinds of tomato organs classification networks are constructed. To reduce overfitting on tomato organ image dataset, before using the dataset to train the individual network, we use many data augmentation techniques to enlarge the dataset such as label-preserving transformations, specifically image rotation at random angles, extracting random patches of one image, image flip horizontal, and altering the intensities of the RGB (red, green, blue) channels of training images. At the same time, the median filter is used to denoise image samples. We train our models end to end using stochastic gradient descent with input image size of 64×64 or 128×128 pixels, mini-batch size of 128 examples, and momentum of 0.9. The classification test results show that the 10 kinds of networks all have very good classification performance by top-1 error rate. Among the 10 kinds of networks, the 8-layer structure network's top-1 error rate is 0.297%, which has the best classification performance. The results reveal that 8-layer structure network can satisfy the features extraction and classification requirements of tomato image dataset with fewer categories. Considering the classification performance and computing speed, referencing Fast R-CNN, with the 8-layer structure network as the basic architecture, and using the region of interest pooling (RoI pooling) layer to replace the last maximum pooling (max-pool) layer, a tomato main organ detector is designed, and named as TD-Net. The detector can output the category probabilities for each object proposal through the soft-max layer. The selective search algorithm is adopted for generating a large number of proposal regions for each tomato plant image, and the non-maximum suppression algorithm is used to suppress the outputs of the TD-Net. Using plant images and proposal regions as input, the detector is trained iteratively. The detection experiment results show that the detection average precisions (AP) of the DCNN-based tomato organ detector for the fruit, flower and stem are 81.64%, 84.48% and 53.94% correspondingly, and the detector can effectively identify the fruits of different maturity degree and the flowers of different flower age. On the same tomato plant image dataset, compared with R-CNN and Fast R-CNN, TD-Net has a better detection performance and detection speed, which illustrates that the method of this paper is effective. [ABSTRACT FROM AUTHOR]

Copyright of Acta Automatica Sinica is the property of Chinese Academy of Sciences, Institute of Automation 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 Acta Automatica Sinica is the property of Chinese Academy of Sciences, Institute of Automation 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 that the features in the process of strip surface defect identification can not be automatically and accurately extracted. In this paper, a new approach to strip surface defect recognition based on deep learning is proposed. This paper first analyzes the basic theory of deep learning, on the basis of this, the basic model of strip surface recognition is established. Then the basic model parameters are trained by the prepared training images. The model extracts image features layer by layer through hidden layers,thus automatically acquire the essential characteristics of the target. Identification and classification are performed. Finally, the validity of the model is verified by testing image. The experimental results show that the accuracy of strip surface defect identification can reach more than 98%, it meets the requirements of strip steel identification. [ABSTRACT FROM AUTHOR]

With the increase of training data and the improvement of the performance of computers, the CNN-based object detection breaks the bottleneck of traditional object detection and has been the main method of current object detection. Therefore, it is a significant research that how to effectively utilize CNN for object detection. Firstly, this paper reviewed how to solve the problems of traditional object detection by CNN. Secondly, it introduced the basic architecture of CNN and described the current research development and widely-used CNN. Thirdly, this paper mainly analyzed and discussed two kinds of ideas and methods of the CNN-based object detection and pointed out the present deficiency. Finally, it concluded the CNN-based object detection and the future direction. [ABSTRACT FROM AUTHOR]

Plant leaf diseases are a serious problem in agricultural production. To solve this problem and prevent diseases deterioration, accurate identification of diseases types is of great significance. In this paper, we proposed a recognition model of plant leaf diseases based on convolutional neural network (CNN), which combines the batch normalization and global pooling methods. The parameters of the traditional CNN model are large and have difficulty to converge. The proposed model was modified in the traditional structure of the CNN, which could optimize the training time and achieve the higher accuracy, and also reduce the size of model. In order to speed up the training convergence, we used the batch normalization layers. We put the input of every convolutional layer in batch, calculated the mean and variance of the batch, and then normalized this batch. We reduced some feature maps of some layers and removed the last full connect layer, with the global pooling layer instead. The proposed model has 5 convolutional layers and 4 pooling layers. In the last pooling layer pool5, the same kernel size of convolutional layer Conv5 was used to take advantage of the information of Conv5's feature map comprehensively. For the image preprocessing, we had zoomed, flipped and rotated the original pictures of dataset randomly to get the augmented dataset, and used the 80% of pictures as the train dataset and the rest as the test dataset. These pictures were quantized to 256×256 dpi for CNN training, and the original dataset and augmented dataset were used to train models. To look for the best size of the first layer kernel, in the first convolutional layer, different kernel sizes i.e. 11×11, 9×9 and 7×7 dpi were used respectively. Furthermore, we chose the type of global pooling layer, like max pooling and average pooling. Then we designed 8 models with different Conv1 kernel sizes or global pooling types. To further improve the efficiency of this model, besides using the Gaussian initialization, we used the other common type of convolutional initialization such as Xavier initialization, and also used the PRelu activation function for each convolution layer. So the optimal model could be selected to recognize the 26 kinds of leaf diseases which involved 14 kinds of plants, and then we analyzed the model's convergence rate, memory usage and robustness. After the experiment, we compared the test accuracy between the traditional model and the proposed model based on original dataset and augmented dataset. The proposed model could accelerate the training convergence, and the test accuracy could achieve about 90% while the traditional model was only about 77% after 3 training epochs. Different kernel sizes of Conv1 had little impact on the accuracy but small kernel was proved to be more beneficial to the recognition of plant diseases, which could get more texture features than the big kernel size filter, and average pooling also made better results than max pooling. We got the best performance model which used the 9×9 dpi kernel size and global average pooling layer. To show the proposed model's performance, we tested the accuracy on each class, and the mean accuracy of augmented test dataset was 99.56%, and the weighted average score of recall and precision rate achieved 99.41%. The proposed model had the size of only 2.6 MB. In addition, compared with the traditional methods, the change of the spatial position of the pictures had little effect on the performance of the improved model, and the proposed model could identify different diseases of various plant leaves. The results show that the model has higher recognition accuracy and stronger robustness, and can be used for the identification of plant leaf diseases. [ABSTRACT FROM AUTHOR]

Copyright of China Sciencepaper is the property of China Sciencepaper 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 classification accuracy of the traditional visual features can not meet the requirements in the application of modern military affairs due to the extremely high similarity of the different objects in the battlefield and low recognition rate in complex conditions. This paper presented a new architecture of the convolutional neural network (CNN) based on PCA whitening for solving the classification problem of large-scale images which contained some specific military objects. It efficiently eliminated the relativity of sample data. enhanced the learning ability and improved the accuracy of the object recognition. It tested and evaluated the new CNN model with the large-scale data from military images and compared with the traditional methods. The experiment results show that the algorithm has higher recognition rate on military object recognition. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Guangdong University of Technology is the property of Journal of Guangdong University of 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.)

This paper aims to solve the problem of high throughput automated data analysis of brain microstructure reconstruction by using the solution of systems engineering that consists of hardware, software and algorithm. To solve the key issues of limit analysis efficiency such as registration of large-volume distorted image, dense neural structure recognition, ambiguous neural connections and multi-source data fusion etc, classification method in accordance with the biological characteristics of neural structure is proposed by the realization of the samples preparation, automatic cutting, microscopic imaging, 3D reconstruction and software platform building etc. Furthermore, it provides a theoretical basis and solution for building engineering platform for large volume neural structure reconstruction and meets the brain science research of high-throughput neural circuit network reconstruction demand. [ABSTRACT FROM AUTHOR]

The utilization of smart phone’ acceleration sensors to identify human behavior is a big topic in intelligence field. Traditional identification methods, such as Bayes, speed learning and decision tree, must first collected acquisition frequency domain features of acceleration sensor data and extracted preferred features. This paper utilized convolutional neural network algorithm to extract smart phone’ accelerate data in three dimensions, and then automatically found out patterns from the data. Finally it used decision tree to identify human behavior from the pattern. Experimental results show that recognition accuracy improves 1.1% to 5.2% in comparison with traditional machine learning methods, especially data set on the large -scale. [ABSTRACT FROM AUTHOR]

For the problem of mobile user activity model over-fitting which leads to the generalization of the problem, this paper put forward an improved mobile user activity recognition method of Dropout DBN (deep belief network). The method randomly changed the probability parameters in algorithms of Dropout, reduced the number of hidden units of network nodes, optimized the network weight in each training. This method improved the accuracy of recognition and generalization when the number of sample decreases. Experimental results show that the average recognition accuracy rate for five activities walking, running, upstairs and downstairs reaches 94. 23% by using random Dropout network, recognition accuracy improves 4. 57% . [ABSTRACT FROM AUTHOR]

To solve the problem that the loss of image feature information and improve the accuracy of image retrieval, when the convolutional neural network (CNN) was used to extract the feature information of the image, this paper proposed an image retrieval algorithm based on improved convolutional neural network LeNet-L. First, it improved LeNet-5 convolution neural network structure, increased depth of network structure. Then, it pre-trained the deep convolutional neural network LeNet-L to extract the high-level semantic features. At last, it obtained the similar images by distance function between the image being retrieved and the one in image database. In Corel dataset, compared with the original model method and the traditional image retrieval method based on SVM and active learning, the proposed method had a higher accuracy. The experimental results show that the improved convolutional neural network has a better retrieval effect. [ABSTRACT FROM AUTHOR]

Deep learning method has very excellent ability of image feature extraction. In order to get rid of disadvantages of traditional methods require a priori knowledge, this paper proposed an image feature extraction algorithm based on the fusion AutoEncoder and convolutional neural network (CNN ). The method introduced a fast sparsity control technique to AutoEncod-er and utilized AutoEncoder to train the basic elements of image and initialized the convolution kernel of CNN. Meanwhile, the algorithm added filtering mechanism to the CNN to keep the sparsity of output characteristics. The results of experiments point out that this method achieves good performance on the Minist handwritten digital library and the Yale face database. Furthermore, the advanced experimental outcomes indicate that the feature extraction model included the filtering technique is more effective than the model without filtering mechanism by using cross-validation with T test. [ABSTRACT FROM AUTHOR]

Copyright of Computer Measurement & Control is the property of Magazine Agency of Computer Measurement & Control 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 Hebei University of Science & Technology is the property of Hebei University of Science & Technology, Journal of Hebei University of 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.)

Copyright of Journal of Image & Graphics is the property of Editorial Office of Journal of Image & Graphics 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.)

With the rapid development of deep learning method in artificial intelligence area, the demand of realizing deep learning applications on HPC platform in the university is also increased. In order to provide computing resources for deep learning users, π supercomputer in Shanghai Jiao Tong University is equipped with 130 CPU cards, and several deep learning frameworks such as TensorFlow are deployed on it. This paper studies to deploy deep learning frameworks on traditional HPC platform and compares the performances of training an Inception model by TensorFlow with GPU on π supercomputer and GPU workstation. The experiment results show that the performance of training on π supercomputer is close to that on NVIDIA Minsky Workstation. [ABSTRACT FROM AUTHOR]

Copyright of Computer Measurement & Control is the property of Magazine Agency of Computer Measurement & Control 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.)