TensorFlow Deep Learning Framework Research Articles

Vision detection and path planning of mobile robots for rebar binding

AbstractFocused on the problems of cumbersome operation, low efficiency, and high cost in the traditional manual rebar binding process, we propose a mobile robot vision detection and path‐planning method for rebar binding to realize automated rebar binding by combining deep learning and path‐planning technology. A MobileNetV3‐SSD rebar binding crosspoints recognition model is built based on TensorFlow deep learning framework, and a crosspoints localization method combining control factor α and feature projection curve is introduced to achieve the localization of unbound crosspoints. In addition, A back‐and‐forth path‐planning algorithm with priority constraints combined with dead zone escape algorithm based on improved A* is proposed to achieve complete coverage path planning of the working area and path transfer of the dead zone. In the field test of the robot prototype, the classification accuracy and localization accuracy reached 94.40% and 90.49%, and the robot was able to reach complete coverage path planning successfully. The experimental results show that the visual detection method can achieve fast, noncontact and intelligent recognition of rebar binding crosspoints, which has good robustness and application value. At the same time, the proposed path‐planning method has higher efficiency in the execution of robot complete coverage path planning, and meets the basic requirements of path planning for rebar binding process.

AUTOMATED DETECTION OF VITRITIS USING ULTRAWIDE-FIELD FUNDUS PHOTOGRAPHS AND DEEP LEARNING.

Evaluate the performance of a deep learning algorithm for the automated detection and grading of vitritis on ultrawide-field imaging. Cross-sectional noninterventional study. Ultrawide-field fundus retinophotographs of uveitis patients were used. Vitreous haze was defined according to the six steps of the Standardization of Uveitis Nomenclature classification. The deep learning framework TensorFlow and the DenseNet121 convolutional neural network were used to perform the classification task. The best fitted model was tested in a validation study. One thousand one hundred eighty-one images were included. The performance of the model for the detection of vitritis was good with a sensitivity of 91%, a specificity of 89%, an accuracy of 0.90, and an area under the receiver operating characteristics curve of 0.97. When used on an external set of images, the accuracy for the detection of vitritis was 0.78. The accuracy to classify vitritis in one of the six Standardization of Uveitis Nomenclature grades was limited (0.61) but improved to 0.75 when the grades were grouped into three categories. When accepting an error of one grade, the accuracy for the six-class classification increased to 0.90, suggesting the need for a larger sample to improve the model performances. A new deep learning model based on ultrawide-field fundus imaging that produces an efficient tool for the detection of vitritis was described. The performance of the model for the grading into three categories of increasing vitritis severity was acceptable. The performance for the six-class grading of vitritis was limited but can probably be improved with a larger set of images.

Construction of Transformer Fault Diagnosis and Prediction Model Based on Deep Learning

The current intelligent diagnosis and prediction methods for transformer faults are prone to low diagnostic accuracy and insufficient trend prediction ability when the fault categories are imbalanced. Therefore, a fault diagnosis and prediction model for transformers was constructed using a deep learning framework. The fault diagnosis model was constructed using a focus loss stack sparse noise reduction autoencoder on the deep learning framework. The prediction model was constructed by fusing long and short term memory networks on the basis of tree structure Parzen optimization, and the two models were validated. The results obtained through validation of the diagnostic model indicate that, when the actual hidden layer is set to 3 and the quantity of neurons is 58, the model accuracy during training and testing reaches 97.5% and 92.5% respectively. After adding 0.001 times the Gaussian white noise, the model accuracy was significantly lifted, so this study set the Gaussian noise coefficient to 0.001. In the comparison with baseline models, the actual classification ability of the research model samples is strong, significantly improving the fault diagnosis ability. In the validation of the prediction model, the three error index values of the research model in the single prediction step of CH4 concentration were 0.0699, 0.0540, and 0.8481%, respectively, and proved to be were lower than in the case of the baseline model. The three error values in the two-step prediction are 0.0194, 0.0161, and 0.6535%, which are also lower than in case of the baseline model. Overall, the diagnosis and prediction model proposed in this paper can provide real-time future numerical predictions of dissolved gas analysis and monitoring data in transformer oil. Furthermore, the research outilnes the future development trend of monitoring and measurement through application of tensor flow deep learning framework in transformer fault diagnosis. The attained prediction results are innovative, and could well complete the purpose of actual transformer fault diagnosis and early warning.

A novel multi-layer stacking ensemble wind power prediction model under Tensorflow deep learning framework considering feature enhancement and data hierarchy processing

Wind power prediction is crucial for energy production, but due to the complicated data characteristics of wind farms, it's difficult to accurately predict wind power output and it is challenging for a single prediction model to properly handle multi-featured time-series data sets. To solve the above problem, the density-based spatial clustering of applications with noise (DBSCAN) method was used to find outliers of data sets, and the recursive feature elimination (RFE) method was used to screen out main features according to importance. Sequentially a multi-layer stacking ensemble learning prediction model was constructed based on data hierarchy processing and feature enhancement techniques. Seven groups of ablation experiments were set to verify the model's effectiveness with the annual time-series data. The experimental results show that the DBSCAN method can effectively find outliers in wind data sets and improve the prediction accuracy and the RFE method can significantly reduce the computing time and improves the generalization ability and prediction accuracy. The multi-layer stacking ensemble model can slow down the overfitting trend of a single prediction model, effectively avoid the model from falling into the local search state, and obtain a strong learner with better generalization ability.

A Multi-Task Fusion Strategy-Based Decision-Making and Planning Method for Autonomous Driving Vehicles.

The autonomous driving technology based on deep reinforcement learning (DRL) has been confirmed as one of the most cutting-edge research fields worldwide. The agent is enabled to achieve the goal of making independent decisions by interacting with the environment and learning driving strategies based on the feedback from the environment. This technology has been widely used in end-to-end driving tasks. However, this field faces several challenges. First, developing real vehicles is expensive, time-consuming, and risky. To further expedite the testing, verification, and iteration of end-to-end deep reinforcement learning algorithms, a joint simulation development and validation platform was designed and implemented in this study based on VTD-CarSim and the Tensorflow deep learning framework, and research work was conducted based on this platform. Second, sparse reward signals can cause problems (e.g., a low-sample learning rate). It is imperative for the agent to be capable of navigating in an unfamiliar environment and driving safely under a wide variety of weather or lighting conditions. To address the problem of poor generalization ability of the agent to unknown scenarios, a deep deterministic policy gradient (DDPG) decision-making and planning method was proposed in this study in accordance with a multi-task fusion strategy. The main task based on DRL decision-making planning and the auxiliary task based on image semantic segmentation were cross-fused, and part of the network was shared with the main task to reduce the possibility of model overfitting and improve the generalization ability. As indicated by the experimental results, first, the joint simulation development and validation platform built in this study exhibited prominent versatility. Users were enabled to easily substitute any default module with customized algorithms and verify the effectiveness of new functions in enhancing overall performance using other default modules of the platform. Second, the deep reinforcement learning strategy based on multi-task fusion proposed in this study was competitive. Its performance was better than other DRL algorithms in certain tasks, which improved the generalization ability of the vehicle decision-making planning algorithm.

A flexible and lightweight deep learning weather forecasting model

AbstractNumerical weather prediction is an established weather forecasting technique in which equations describing wind, temperature, pressure and humidity are solved using the current atmospheric state as input. This study examines deep learning to forecast weather given historical data from two London-based locations. Two distinct Bi-LSTM recurrent neural network models were developed in the TensorFlow deep learning framework and trained to make predictions in the next 24 and 72 h, given the past 120 h. The first trained neural network predicted temperature at Kew Gardens with a forecast accuracy of $$\pm$$ ± 2 $${}^{\circ }$$ ∘ C in 73% of instances in a whole unseen year, and a root mean squared errors of 1.45 $${}^{\circ }$$ ∘ C. The second network predicted 72-h air temperature and relative humidity at Heathrow with root mean squared errors 2.26 $${}^{\circ }$$ ∘ C and 14% respectively and 80% of the temperature predictions were within $$\pm$$ ± 3 $${}^{\circ }$$ ∘ C while 80% of relative humidity predictions were within $$\pm$$ ± 20%. Both networks were trained with five years of historical data, with cloud training times of over a minute (24-h network) and three minutes (72-h).

A comprehensive assessment of hurdle and zero-inflated models for single cell RNA-sequencing analysis.

Single cell RNA-sequencing (scRNA-seq) technology has significantly advanced the understanding of transcriptomic signatures. Although various statistical models have been used to describe the distribution of gene expression across cells, a comprehensive assessment of the different models is missing. Moreover, the growing number of features associated with scRNA-seq datasets creates new challenges for analytical accuracy and computing speed. Here, we developed a Python-based package (TensorZINB) to solve the zero-inflated negative binomial (ZINB) model using the TensorFlow deep learning framework. We used a sequential initialization method to solve the numerical stability issues associated with hurdle and zero-inflated models. A recursive feature selection protocol was used to optimize feature selections for data processing and downstream differentially expressed gene (DEG) analysis. We proposed a class of hybrid models combining nested models to further improve the model's performance. Additionally, we developed a new method to convert a continuous distribution to its equivalent discrete form, so that statistical models can be fairly compared. Finally, we showed that the proposed TensorFlow algorithm (TensorZINB) was numerically stable and that its computing speed and performance were superior to those of existing ZINB solvers. Moreover, we implemented seven hurdle and zero-inflated statistical models in Python and systematically assessed their performance using a real scRNA-seq dataset. We demonstrated that the ZINB model achieved the lowest Akaike information criterion compared with other models tested. Taken together, TensorZINB was accurate, efficient and scalable for the implementation of ZINB and for large-scale scRNA-seq data analysis with DEG identification.

Impact of Data Augmentation Techniques on the Implementation of a Combination Model of Convolutional Neural Network (CNN) and Multilayer Perceptron (MLP) for the Detection of Diseases in Rice Plants

Detection of disease in rice plants is important to avoid damage and reduce yields. In this study, the influence of data augmentation techniques on the application of a combination model of Convolutional Neural Network (CNN) and Multilayer Perceptron (MLP) was carried out to detect diseases in rice plants. This study uses a dataset of rice images for disease detection which contains 6034 images of rice with five categories, namely Bacterial Leaf Bligh, Blast, Brown Spot, normal, and tungro. This dataset is divided into three parts, namely training data, validation data and test data. The data augmentation techniques used are rotation, brightness and zoom on rice images. The combination model of CNN and MLP is built using the Python programming language and the TensorFlow deep learning framework. In measuring the success rate of the built model, it can be measured using the accuracy, precision and recall values obtained in the model test. Several scenarios were carried out to produce the best model, namely the use of data augmentation techniques, the number of layers and the number of iterations (epochs). From the experiments that have been carried out which have been tested with data as many as 25 digital images, the best model is obtained with an testing accuracy of 92%, 94% precision and 92% recall. This model applies a random zoom augmentation technique with a value of 0.5 – 1.0, CNN+MLP with 3 layers and a dataset ratio of 80:20 and an epoch early stop, This result has increased by more than 10%.

Sentiment Analysis of User Comment Text based on LSTM

Taking the user-generated Chinese comment dataset on online platforms as the research object, we constructed word2vec word vectors using gensim and built a sentiment analysis model based on LSTM using the TensorFlow deep learning framework. From the perspective of mining user comment data on the platform, we analyzed the sentiment tendency of user comments, providing data support for hotels to understand consumers' real sentiment tendencies and improve their own service quality. Through analysis of the validation dataset results obtained by crawling the website, the accuracy of this LSTM model can reach up to 0.89, but there is still much room for improvement in the accuracy of sentiment analysis for some datasets. In future research, this model needs further optimization to obtain a stable and more accurate deep-learning model.

Image segmentation for blind lanes based on improved SegNet model

The position of blind lanes must be correctly determined in order for blind people to travel safely. Aiming at the low accuracy and slow speed of traditional blind lanes image segmentation algorithms, a semantic segmentation method based on SegNet and MobileNetV3 is proposed. The main idea is to replace the coding part of the original SegNet model with the feature extraction part of MobileNetV3 and remove the pooling layer. Blind lanes images were collected through online search and self-shooting, and then the data were manually marked by LabelMe software and trained on TensorFlow deep learning framework. The experimental results show that the improved model has high segmentation accuracy and recognition speed. The pixel accuracy of blind lanes segmentation is 98.21%, the mean intersection over union is 96.29%, and the average time for processing a 416 × 416 image is 0.057 s, which meets the real-time requirements of the blind guidance system.

The safety evaluation method of steel enterprises based on deep learning

As an important basic industry of national economy, the iron and steel industry has provided an important raw material guarantee for a long time. However there are a large number of hazard sources which are prone to safety accidents in the production process. Then safety evaluation in the production system is highly needed to effectively prevent the occurrence of accidents in iron and steel enterprises. Hence we introduce a method based on deep learning model to evaluate safety of the enterprises. Firstly, the risk factors and casualties in production process are investigated, and a set of safety evaluation index system is constructed.Secondly, since deep neural network model has the characteristics of strong feature extraction ability and simple model structure, we design a safety evaluation model based on deep neural network. The 25-dimensional evaluation index value is the input of the network, and the network output corresponds to five risk levels. On this basis, the optimization algorithm of deep neural network model is designed to explore the mapping relationship between risk characteristics and safety level. Tensorflow deep learning framework is used to build the evaluation model, classification loss function and network optimization method are designed to train the model. Finally, through experiments, the optimal model structure is determined by comparing the influence of different parameter optimization strategies, different hidden layer structures, and different activation functions on the safety evaluation performance. A three hidden layer structure with the Adam back propagation algorithm and LeakyRelu activation function is adopted to obtain higher accuracy and faster convergence rate. The experiments show that our evaluation model provides an operational method for evaluating the safety management status of iron and steel enterprises.

Differentiating Magnetic Resonance Images of Pyogenic Spondylitis and Spinal Modic Change Using a Convolutional Neural Network.

A retrospective analysis of magnetic resonance imaging (MRI). The study aimed to evaluate the performance of a convolutional neural network (CNN) to differentiate spinal pyogenic spondylitis from Modic change on MRI. We compared the performance of CNN to that of four clinicians. Discrimination between pyogenic spondylitis and spinal Modic change is crucial in clinical practice. CNN deep-learning approaches for medical imaging are being increasingly utilized. We retrospectively reviewed MRIs from pyogenic spondylitis and spinal Modic change patients. There were 50 patients per group. Sagittal T1-weighted (T1WI), sagittal T2-weighted (T2WI), and short TI inversion recovery (STIR) MRIs were used for CNN training and validation. The deep learning framework Tensorflow was used to construct the CNN architecture. To evaluate CNN performance, we plotted the receiver operating characteristic curve and calculated the area under the curve. We compared the accuracy, sensitivity, and specificity of CNN diagnosis to that of a radiologist, spine surgeon, and two orthopedic surgeons. The CNN-based area under the curves of the receiver operating characteristic curve from the T1WI, T2WI, and STIR were 0.95, 0.94, and 0.95, respectively. The accuracy of the CNN was significantly greater than that of the four clinicians on T1WI and STIR (P<0.05), and better than a radiologist and one orthopedic surgeon on the T2WI (P<0.05). The sensitivity was significantly better than that of the four clincians on T1WI and STIR (P<0.05), and better than a radiologist and one orthopedic surgeon on the T2WI (P<0.05). The specificity was significantly better than one orthopedic surgeon on T1WI and T2WI (P<0.05) and better than both orthopedic surgeons on STIR (P<0.05). We differentiated between Modic changes and pyogenic spondylitis using a CNN that interprets MRI. The performance of the CNN was comparable to, or better than, that of the four clinicians.

Estimation of Fusarium Head Blight Severity Based on Transfer Learning

The recognition accuracy of traditional image recognition methods is heavily dependent on the design of complicated and tedious hand-crafted features. In view of the problems of poor accuracy and complicated feature extraction, this study presents a methodology for the estimation of the severity of wheat Fusarium head blight (FHB) with a small sample dataset based on transfer learning technology and convolutional neural networks (CNNs). Firstly, we utilized the potent feature learning and feature expression capabilities of CNNs to realize the automatic learning of FHB characteristics. Using transfer learning technology, VGG16, ResNet50, and MobileNetV1 models were pre-trained on the ImageNet. The knowledge was transferred to the estimation of FHB severity, and the fully connected (FC) layer of the models was modified. Secondly, acquiring the wheat images at the peak of the outbreak of FHB as the research object, after preprocessing for size filling on the wheat images, the image dataset was expanded with operations such as mirror flip, rotation transformation, and superimposed noise to improve the performance of the model and reduce the overfitting of models. Finally, under the Tensorflow deep learning framework, the VGG16, ResNet50, and MobileNetV1 models were subjected to transfer learning. The results showed that in the case of transfer learning and data augmentation, the ResNet50 model in Accuracy, Precision, Recall, and F1 score was better than the other two models, giving the highest accuracy of 98.42% and F1 score of 97.86%. The ResNet50 model had the highest recognition accuracy, providing technical support and reference for the accurate recognition of FHB.

Application of efficient recognition algorithm based on deep neural network in English teaching scene

The recognition of English texts in teaching scenes is a practical research direction. English text recognition can be widely used in English teaching scenes, such as assisting teachers to recognise students’ English homework, text positioning before text translation, developing outdoor classrooms, assisting junior students in scene understanding and so on. To identify English information in different scenes as accurately as possible, identifying the corresponding text content is the key. Based on a deep neural network, this paper proposes GCN-Attention English recognition algorithm. The experiment adopts the deep learning framework Tensorflow, which combines 104 × 104 size GCN with an attention mechanism for training. The output of GCN is used to train the cyclic neural network to continuously predict the next most likely letter in the sequence. The goal of training is to match the output words with the expected words as much as possible. The test results show that the model can have a good recognition accuracy for the scene image data set used in teaching.

Water Leakage and Crack Identification in Tunnels Based on Transfer-Learning and Convolutional Neural Networks

In order to solve the problems of long artificial time consumption, the inability to standardize the degree of damage, and the difficulty of maintaining data in traditional tunnel disease detection methods, this paper proposes the use of Residual Network (ResNet) models for tunnel water leakage and crack detection. ResNet proposes a residual learning framework to ease the training of networks that are deeper than those previously used. Furthermore, ResNet explicitly reformulates the layers as learning the residual functions of the reference layer inputs, rather than learning the unreferenced functions. The ResNet model is built on the Tensorflow Deep Learning (DL) framework and transfer-learning is used to optimize the model. The ResNet-V1 can be obtained by pre-training in ImageNet. The fully connected layers of the ResNet-V1 were modified to four classifications of tunnel disease. Then, the SoftMax function is used to recognize the tunnel diseases. Four network structures have been chosen, i.e., ResNet34 and ResNet50, with and without Transfer-learning, respectively. Those models were selected for testing and training on the sample dataset, and these four network structures were compared and analyzed using five types of evaluation indicators, which are the confusion matrix, accuracy, precision, recall ratio and F1. In identifying tunnel cracks and water leakage, the accuracy of ResNet50 and ResNet34 using the transfer-learning were 96.30% and 91.29%, and the accuracy of ResNet50 was 5.01% higher than that of ResNet34; for the network structure without the transfer-learning, the accuracy of ResNet50 was 90.36% and ResNet34’s accuracy was 87.87%. These data show that the accuracy of ResNet50 is higher than that of ResNet34 with or without the transfer-learning, and the deep structure framework is superior in the identification of tunnel diseases; secondly, comparing the network structures with and without the transfer-learning, it can be found that using the Transfer-Learning can improve the ResNet network’s accuracy for tunnel disease identification. The experiments and reliability analysis demonstrate the intelligent tunnel disease identification method proposed in this paper, and its good robustness and generalization performance. This method can be used for the rapid identification of cracks and water leakage in a tunnel survey, construction and maintenance, which has practical engineering implications for tunnel disease detection.

Counterfeit Currency Detection based on AI

Abstract: The use of technology has grown tremendously within the few years it has made it easier to have access to advanced printing equipment in the industry which resulted in color printing of currencies to produce counterfeit notes across the country. To eliminate such unethical activities of printing counterfeit currency it is mandatory to make a system that detects the fake currency, In systems such as a money exchanger for example ATMs and vending machines, counterfeit currency notes must be detected beforehand exchanging process takes place. In the past, there have been similar systems developed based on methods such as image processing techniques that are done on the Matlab platform and other such platforms these methods possess some limitations including being less efficient and time-consuming. Our system is designed to eliminate all of the above problems through the use of deep learning techniques by detecting the features of currencies and determining whether its fake with a great accuracy rate. our proposed system verifies the Indian currency notes using Deep learning, deep learning helps in extracting meaningful information from the dataset fed into the machine using a set of methods to perform the classification of images. Our project makes use of the deep learning framework TensorFlow and its high-level API Keras which simplifies the creation of the model making it easier to achieve a less time-consuming and accurate model. Index Terms: Currency Detection, Deep Learning, Deep Neural Network, Convolution Neural Network

Land Use Classification Method of Remote Sensing Images for Urban and Rural Planning Monitoring Using Deep Learning

Aiming at the problems that most existing segmentation methods are difficult to deal with the imbalance of remote sensing image distribution and the overlap of segmentation target edges, a land use classification method of remote sensing image for urban and rural planning monitoring based on deep learning is proposed. Firstly, the U-Net is improved by pooling index upsampling and dimension superposition. The improvement can not only extract high-level abstract features but also extract low-level detail features, so as to reduce the loss of image edge information in the process of deconvolution. Then, the batch normalization and scaling exponential linear unit (SeLU) are used to improve the U-Net model. Finally, the improved U-Net model is applied to the classification of remote sensing images of land use types to realize dynamic monitoring. The experimental analysis of the proposed method based on TensorFlow deep learning framework shows that its total accuracy exceeds 94%. The segmentation effect of land use types in remote sensing images is good.

5G Massive MIMO Signal Detection Algorithm Based on Deep Learning.

Aiming at the problems of poor signal detection effect caused by many interference factors in large-scale MIMO technology scene, this paper proposes a 5G massive MIMO signal detection algorithm based on deep learning. Firstly, the MIMO system model based on neural network is constructed, and Deep Neural Network (DNN) detection is introduced into the receiver of the traditional MIMO system to obtain the information bits or codewords and channel state information transmitted by transmitters. Then, the end-to-end training method is adopted to make neural network learn the mapping relationship of information bits or codewords transmitted by system transceivers. Furthermore, DNN detector is improved based on Simplified Message Passing Detection (sMPD) algorithm, and the correction factor is updated continuously to optimize network parameters to realize the accurate detection and decoding of the MIMO system. Finally, the proposed algorithm is experimentally analyzed based on the TensorFlow deep learning framework. Experimental results show that when signal-to-noise ratio is 10 dB, the bit error rate and mean square error are lower than 0.005 and 0.1, respectively.

Application of Particle Swarm Optimization Combined with Long and Short-term Memory Networks for Short-term Load Forecasting

In this paper, we apply the Long Short-Term Memory (LSTM) network to short-term load forecasting, and use the TensorFlow deep learning framework to build a Particle Swarm Optimization (PSO) model to optimize the parameters of the LSTM. Optimization (PSO) model to optimize the parameters of LSTM. In this paper, we use the meteorological data and historical load data of a certain place as the input of LSTM before and after optimization, and compare the model with the BP Neural Network before and after optimization, and the results show that the PSO-LSTM model has higher reliability and prediction accuracy.

Determining the short-term neurological prognosis for acute cervical spinal cord injury using machine learning

It is challenging to predict neurological outcomes of acute spinal cord injury (SCI) considering issues such as spinal shock and injury heterogeneity. Deep learning-based radiomics (DLR) were developed to quantify the radiographic characteristics automatically using a convolutional neural network (CNN), and to potentially allow the prognostic stratification of patients. We aimed to determine the functional prognosis of patients with cervical SCI using machine learning approach based on MRI and to assess the ability to predict the neurological outcomes. We retrospectively analyzed the medical records of SCI patients (n=215) who had undergone MRI and had an American Spinal cord Injury Association Impairment Scale (AIS) assessment at 1 month after injury, enrolled with a total of 294 MR images. Sagittal T2-weighted MR images were used for the CNN training and validation. The deep learning framework TensorFlow was used to construct the CNN architecture. After we calculated the probability of the AIS grade using the DLR, we built the identification model based upon the random forest using 3 features: the probability of each AIS grade obtained by the DLR method, age, and the initial AIS grade at admission. We performed a statistical evaluation between the actual and predicted AIS. The accuracy, precision, recall and f1 score of the ensemble model based on the DLR and RF were 0.714, 0.590, 0.565 and 0.567, respectively. The present study demonstrates that prediction of the short-term neurological outcomes for acute cervical spinal cord injury based on MRI using machine learning is feasible.