Rethinking the Random Cropping Data Augmentation Method Used in the Training of CNN-Based SAR Image Ship Detector

The random cropping data augmentation method is widely used to train convolutional neural network (CNN)-based target detectors to detect targets in optical images (e.g., COCO datasets). It can expand the scale of the dataset dozens of times while consuming only a small amount of calculations when training the neural network detector. In addition, random cropping can also greatly enhance the spatial robustness of the model, because it can make the same target appear in different positions of the sample image. Nowadays, random cropping and random flipping have become the standard configuration for those tasks with limited training data, which makes it natural to introduce them into the training of CNN-based synthetic aperture radar (SAR) image ship detectors. However, in this paper, we show that the introduction of traditional random cropping methods directly in the training of the CNN-based SAR image ship detector may generate a lot of noise in the gradient during back propagation, which hurts the detection performance. In order to eliminate the noise in the training gradient, a simple and effective training method based on feature map mask is proposed. Experiments prove that the proposed method can effectively eliminate the gradient noise introduced by random cropping and significantly improve the detection performance under a variety of evaluation indicators without increasing inference cost.

Download Full-text

Application of ANN Back-Propagation for Fracture Design Parameters in Extra-Low Cycle Rotating Bending Fatigue

Materials Science Forum ◽

10.4028/www.scientific.net/msf.610-613.450 ◽

2009 ◽

Vol 610-613 ◽

pp. 450-453

Author(s):

Hong Yan Duan ◽

You Tang Li ◽

Jin Zhang ◽

Gui Ping He

Keyword(s):

Neural Network ◽

Neural Network Model ◽

Back Propagation ◽

Training Data ◽

Design Parameters ◽

Ann Model ◽

Bending Fatigue ◽

The Neural Network ◽

Rotating Bending Fatigue ◽

Rotating Bending

The fracture problems of ecomaterial (aluminum alloyed cast iron) under extra-low cycle rotating bending fatigue loading were studied using artificial neural networks (ANN) in this paper. The training data were used in the formation of training set of ANN. The ANN model exhibited excellent in results comparison with the experimental results. It was concluded that predicted fracture design parameters by the trained neural network model seem more reasonable compared to approximate methods. It is possible to claim that, ANN is fairly promising prediction technique if properly used. Training ANN model was introduced at first. And then the Training data for the development of the neural network model was obtained from the experiments. The input parameters, notch depth, the presetting deflection and tip radius of the notch, and the output parameters, the cycle times of fracture were used during the network training. The neural network architecture is designed. The ANN model was developed using back propagation architecture with three layers jump connections, where every layer was connected or linked to every previous layer. The number of hidden neurons was determined according to special formula. The performance of system is summarized at last. In order to facilitate the comparisons of predicted values, the error evaluation and mean relative error are obtained. The result show that the training model has good performance, and the experimental data and predicted data from ANN are in good coherence.

Download Full-text

Application of ANN Back-Propagation for Fracture Design Parameters of Middle Carbon Steel in Extra-Low Cycle Bend Torsion Loading

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.345.272 ◽

2013 ◽

Vol 345 ◽

pp. 272-276 ◽

Cited By ~ 1

Author(s):

Hong Yan Duan ◽

You Tang Li ◽

Zhi Jia Sun ◽

Yang Yang Zhang

Keyword(s):

Neural Network ◽

Carbon Steel ◽

Network Model ◽

Neural Network Model ◽

Network Architecture ◽

Back Propagation ◽

Training Data ◽

Design Parameters ◽

Ann Model ◽

The Neural Network

The fracture problems of medium carbon steel (MCS) under extra-low cycle bend torsion loading were studied using artificial neural networks (ANN) in this paper. The training data were used in the formation of training set of ANN. The ANN model exhibited excellent comparison with the experimental results. It was concluded that predicted fracture design parameters by the trained neural network model seem more reasonable compared to approximate methods. It is possible to claim that, ANN is fairly promising prediction technique if properly used. Training ANN model was introduced at first. And then the Training data for the development of the neural network model was obtained from the experiments. The input parameters, notch depth and tip radius of the notch, and the output, the cycle times of fracture were used during the network training. The neural network architecture is designed. The ANN model was developed using back propagation architecture with three layers jump connections, where every layer was connected or linked to every previous layer. The number of hidden neurons was determined according to special formula. The performance of system is summarized at last. In order to facilitate the comparisons of predicted values, the error evaluation and mean relative error are obtained. The result show that the training model has good performance, and the experimental data and predicted data from ANN are in good coherence.

Download Full-text

System Identification of UAV Alap-Alap Using Back Propagation Neural Network

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.373-375.1212 ◽

2013 ◽

Vol 373-375 ◽

pp. 1212-1219

Author(s):

Afrias Sarotama ◽

Benyamin Kusumoputro

Keyword(s):

Neural Network ◽

Back Propagation ◽

Back Propagation Neural Network ◽

Training Data ◽

Data Sets ◽

Input Output ◽

Data Set ◽

Flight Data ◽

The Neural Network ◽

Aerial Vehicle

A good model is necessary in order to design a controller of a system off-line. It is especially beneficial in the implementation of new advanced control schemes in Unmanned Aerial Vehicle (UAV). Considering the safety and benefit of an off-line tuning of the UAV controllers, this paper identifies a dynamic MIMO UAV nonlinear system which is derived based on the collection of input-output data taken from a test flights (36250 samples data). These input-output sample flight data are grouped into two flight data sets. The first flight data set, a chirp signal, is used for training the neural network in order to determine parameters (weights) for the network. Validation of the network is performed using the second data set, which is not used for training, and is a representation of UAV circular flight movement. An artificial neural network is trained using the training data set and thereafter the network is excited by the second set input data set. The predicted outputs based on our proposed Neural Network model is similar to the desired outputs (roll, pitch and yaw) which has been produced by real UAV system.

Download Full-text

Application of ANN Back-Propagation for An Alloy Reinforced Ceramics/Metal Composite under Extra-Low Cycle Bending Fatigue Loading

Advanced Materials Research ◽

10.4028/www.scientific.net/amr.105-106.108 ◽

2010 ◽

Vol 105-106 ◽

pp. 108-111

Author(s):

Zhi Yuan Rui ◽

Hong Yan Duan ◽

Chun Li Lei ◽

Xing Chun Wei

Keyword(s):

Neural Network ◽

Network Model ◽

Neural Network Model ◽

Back Propagation ◽

Propagation Model ◽

Training Data ◽

Design Parameters ◽

Metal Composite ◽

Ann Model ◽

The Neural Network

Artificial neural network (ANN) back-propagation model was developed to predict the fracture design parameters in reinforced ceramic matrix composites (CMCS).Training ANN model was introduced at first. And then the Training data for the development of the neural network model was obtained from the experiments. The input parameters, the presetting deflection and tip radius of the notch, and the output, the cycle times of fracture were used during the network training. The neural network architecture is designed. The ANN model was developed using back propagation architecture with three layers jump connections, where every layer was connected or linked to every previous layer. The number of hidden neurons was determined according to special formula. The performance of system is summarized at last. The ANN model exhibited excellent comparison with the experimental results. It was concluded that predicted fracture design parameters by the trained neural network model seem more reasonable compared to approximate methods. It is possible to claim that, ANN is fairly promising prediction technique if properly used.

Download Full-text

Identification of Anesthesia Stages from EEG Signals using Wavelet Entropy and Backpropagation Neural Network

Global Journal of Computer Science and Technology ◽

10.34257/gjcstdvol19is1pg17 ◽

2019 ◽

pp. 17-20

Author(s):

Ahmed Abdal Shafi Rasel

Keyword(s):

Neural Network ◽

Back Propagation ◽

Back Propagation Neural Network ◽

Training Data ◽

Eeg Signals ◽

The Neural Network ◽

Power Spectral ◽

Eeg Data ◽

Density Analysis ◽

Key Features

This study focuses on entropy based analysis of EEG signals for extracting features for a neural network based solution for identifying anesthetic levels. The process involves an optimized back propagation neural network with a supervised learning method. We provided the extracted features from EEG signals as training data for the neural network. The target outputs provided are levels of anesthesia stages. Wavelet analysis provides more effective extraction of key features from EEG data than power spectral density analysis using Fourier transform. The key features are used to train the Back Propagation Neural Network (BPNN) for pattern classification network. The final result shows that entropy-based feature extraction is an effective procedure for classifying EEG data.

Download Full-text

Automated Ventricular System Segmentation in Paediatric Patients Treated for Hydrocephalus Using Deep Learning Methods

BioMed Research International ◽

10.1155/2019/3059170 ◽

2019 ◽

Vol 2019 ◽

pp. 1-9 ◽

Cited By ~ 2

Author(s):

Michał Klimont ◽

Mateusz Flieger ◽

Jacek Rzeszutek ◽

Joanna Stachera ◽

Aleksandra Zakrzewska ◽

...

Keyword(s):

Neural Network ◽

Data Augmentation ◽

Ct Images ◽

Policy Transfer ◽

Training Data ◽

Intraobserver Variability ◽

Practical Applications ◽

Brain Scans ◽

Rate Policy ◽

Ct Brain

Hydrocephalus is a common neurological condition that can have traumatic ramifications and can be lethal without treatment. Nowadays, during therapy radiologists have to spend a vast amount of time assessing the volume of cerebrospinal fluid (CSF) by manual segmentation on Computed Tomography (CT) images. Further, some of the segmentations are prone to radiologist bias and high intraobserver variability. To improve this, researchers are exploring methods to automate the process, which would enable faster and more unbiased results. In this study, we propose the application of U-Net convolutional neural network in order to automatically segment CT brain scans for location of CSF. U-Net is a neural network that has proven to be successful for various interdisciplinary segmentation tasks. We optimised training using state of the art methods, including “1cycle” learning rate policy, transfer learning, generalized dice loss function, mixed float precision, self-attention, and data augmentation. Even though the study was performed using a limited amount of data (80 CT images), our experiment has shown near human-level performance. We managed to achieve a 0.917 mean dice score with 0.0352 standard deviation on cross validation across the training data and a 0.9506 mean dice score on a separate test set. To our knowledge, these results are better than any known method for CSF segmentation in hydrocephalic patients, and thus, it is promising for potential practical applications.

Download Full-text

Hysteresis Modeling and Compensation of Fast Steering Mirrors with Hysteresis Operator Based Back Propagation Neural Networks

Micromachines ◽

10.3390/mi12070732 ◽

2021 ◽

Vol 12 (7) ◽

pp. 732

Author(s):

Kairui Cao ◽

Guanglu Hao ◽

Qingfeng Liu ◽

Liying Tan ◽

Jing Ma

Keyword(s):

Neural Network ◽

Back Propagation ◽

Hysteresis Model ◽

Cooperative Movement ◽

Hysteresis Operator ◽

The Neural Network ◽

Hysteresis Nonlinearity ◽

Hysteresis Modeling ◽

Hysteresis Characteristics ◽

Inverse Hysteresis

Fast steering mirrors (FSMs), driven by piezoelectric ceramics, are usually used as actuators for high-precision beam control. A FSM generally contains four ceramics that are distributed in a crisscross pattern. The cooperative movement of the two ceramics along one radial direction generates the deflection of the FSM in the same orientation. Unlike the hysteresis nonlinearity of a single piezoelectric ceramic, which is symmetric or asymmetric, the FSM exhibits complex hysteresis characteristics. In this paper, a systematic way of modeling the hysteresis nonlinearity of FSMs is proposed using a Madelung’s rules based symmetric hysteresis operator with a cascaded neural network. The hysteresis operator provides a basic hysteresis motion for the FSM. The neural network modifies the basic hysteresis motion to accurately describe the hysteresis nonlinearity of FSMs. The wiping-out and congruency properties of the proposed method are also analyzed. Moreover, the inverse hysteresis model is constructed to reduce the hysteresis nonlinearity of FSMs. The effectiveness of the presented model is validated by experimental results.

Download Full-text

Robust Approach to Supervised Deep Neural Network Training for Real-Time Object Classification in Cluttered Indoor Environment

Applied Sciences ◽

10.3390/app11157148 ◽

2021 ◽

Vol 11 (15) ◽

pp. 7148

Author(s):

Bedada Endale ◽

Abera Tullu ◽

Hayoung Shi ◽

Beom-Soo Kang

Keyword(s):

Neural Network ◽

Deep Learning ◽

Real Time ◽

Network Architecture ◽

Input Data ◽

Deep Neural Network ◽

Data Augmentation ◽

Object Classification ◽

Training Data ◽

Gradient Descent Algorithm

Unmanned aerial vehicles (UAVs) are being widely utilized for various missions: in both civilian and military sectors. Many of these missions demand UAVs to acquire artificial intelligence about the environments they are navigating in. This perception can be realized by training a computing machine to classify objects in the environment. One of the well known machine training approaches is supervised deep learning, which enables a machine to classify objects. However, supervised deep learning comes with huge sacrifice in terms of time and computational resources. Collecting big input data, pre-training processes, such as labeling training data, and the need for a high performance computer for training are some of the challenges that supervised deep learning poses. To address these setbacks, this study proposes mission specific input data augmentation techniques and the design of light-weight deep neural network architecture that is capable of real-time object classification. Semi-direct visual odometry (SVO) data of augmented images are used to train the network for object classification. Ten classes of 10,000 different images in each class were used as input data where 80% were for training the network and the remaining 20% were used for network validation. For the optimization of the designed deep neural network, a sequential gradient descent algorithm was implemented. This algorithm has the advantage of handling redundancy in the data more efficiently than other algorithms.

Download Full-text

Gravity Control-Based Data Augmentation Technique for Improving VR User Activity Recognition

Symmetry ◽

10.3390/sym13050845 ◽

2021 ◽

Vol 13 (5) ◽

pp. 845

Author(s):

Dongheun Han ◽

Chulwoo Lee ◽

Hyeongyeop Kang

Keyword(s):

Activity Recognition ◽

Large Scale ◽

Data Augmentation ◽

Training Data ◽

Measurement Unit ◽

Gravitational Acceleration ◽

The Neural Network ◽

Typical Data ◽

Robust Recognition ◽

Gravity Acceleration

The neural-network-based human activity recognition (HAR) technique is being increasingly used for activity recognition in virtual reality (VR) users. The major issue of a such technique is the collection large-scale training datasets which are key for deriving a robust recognition model. However, collecting large-scale data is a costly and time-consuming process. Furthermore, increasing the number of activities to be classified will require a much larger number of training datasets. Since training the model with a sparse dataset can only provide limited features to recognition models, it can cause problems such as overfitting and suboptimal results. In this paper, we present a data augmentation technique named gravity control-based augmentation (GCDA) to alleviate the sparse data problem by generating new training data based on the existing data. The benefits of the symmetrical structure of the data are that it increased the number of data while preserving the properties of the data. The core concept of GCDA is two-fold: (1) decomposing the acceleration data obtained from the inertial measurement unit (IMU) into zero-gravity acceleration and gravitational acceleration, and augmenting them separately, and (2) exploiting gravity as a directional feature and controlling it to augment training datasets. Through the comparative evaluations, we validated that the application of GCDA to training datasets showed a larger improvement in classification accuracy (96.39%) compared to the typical data augmentation methods (92.29%) applied and those that did not apply the augmentation method (85.21%).

Download Full-text

Safety Warning Model of Coal Face Based on FCM Fuzzy Clustering and GA-BP Neural Network

Symmetry ◽

10.3390/sym13061082 ◽

2021 ◽

Vol 13 (6) ◽

pp. 1082

Author(s):

Fanqiang Meng

Keyword(s):

Neural Network ◽

Genetic Algorithm ◽

Early Warning ◽

Back Propagation ◽

Back Propagation Neural Network ◽

Coal Face ◽

Early Warning Model ◽

Working Face ◽

The Neural Network ◽

Warning Model

Risk and security are two symmetric descriptions of the uncertainty of the same system. If the risk early warning is carried out in time, the security capability of the system can be improved. A safety early warning model based on fuzzy c-means clustering (FCM) and back-propagation neural network was established, and a genetic algorithm was introduced to optimize the connection weight and other properties of the neural network, so as to construct the safety early warning system of coal mining face. The system was applied in a coal face in Shandong, China, with 46 groups of data as samples. Firstly, the original data were clustered by FCM, the input space was fuzzy divided, and the samples were clustered into three categories. Then, the clustered data was used as the input of the neural network for training and prediction. The back-propagation neural network and genetic algorithm optimization neural network were trained and verified many times. The results show that the early warning model can realize the prediction and early warning of the safety condition of the working face, and the performance of the neural network model optimized by genetic algorithm is better than the traditional back-propagation artificial neural network model, with higher prediction accuracy and convergence speed. The established early warning model and method can provide reference and basis for the prediction, early warning and risk management of coal mine production safety, so as to discover the hidden danger of working face accident as soon as possible, eliminate the hidden danger in time and reduce the accident probability to the maximum extent.

Download Full-text