Improving Convolutional Neural Network (CNN) Architecture (miniVGGNet) with Batch Normalization and Learning Rate Decay Factor for Image Classification (original) (raw)
Related papers
An Empirical Study on Position of the Batch Normalization Layer in Convolutional Neural Networks
2019 5th Iranian Conference on Signal Processing and Intelligent Systems (ICSPIS)
In this paper, we have studied how training of the convolutional neural networks (CNNs) can be affected by changing the position of the batch normalization (BN) layer. Three different convolutional neural networks have been chosen for our experiments. These networks are AlexNet, VGG-16, and ResNet-20. We show that the speed-up provided by the BN algorithm can be further improved by using the BN in positions other than the one suggested by its original paper. Also, we discuss how the BN layer in a certain position can aid the training of one network but not the other. Three different positions for the BN layer have been studied in this research, these positions are: BN layer between the convolution layer and the non-linear activation function, BN layer after the non-linear activation function and finally, the BN layer before each of the convolutional layers.
Classification of MNIST Image Dataset Using Improved Convolutional Neural Network
International Journal for Research in Applied Science & Engineering Technology (IJRASET), 2022
Convolutional Neural Network (CNN) holds the current research interest in the ever-evolving image classification field. Accurate classifying the image data with minimum of time is highly desired. But the traditional CNN architecture often fails to generate the appropriate outcome for large dataset. So, a modified approach of CNN is proposed here which is the combination of data augmentation and batch normalization embedded with CNN. Now a days identifying or classifying digits accurately with variety of modes is really a task of challenge. The advantages of the proposed approach are noted when it is applied to the popular MNIST dataset used for digit classification. The proposed approach has been compared with some existing techniques and results infer that the validation, training loss and testing accuracy of the proposed approach are more superior as compared to the state-of-art approaches.
An Optimized Architecture of Image Classification Using Convolutional Neural Network
International Journal of Image, Graphics and Signal Processing, 2019
The convolutional neural network (CNN) is the type of deep neural networks which has been widely used in visual recognition. Over the years, CNN has gained lots of attention due to its high capability to appropriately classifying the images and feature learning. However, there are many factors such as the number of layers and their depth, number of features map, kernel size, batch size, etc. They must be analyzed to determine how they influence the performance of network. In this paper, the performance evaluation of CNN is conducted by designing a simple architecture for image classification. We evaluated the performance of our proposed network on the most famous image repository name CIFAR-10 used for the detection and classification task. The experiment results show that the proposed network yields the best classification accuracy as compared to existing techniques. Besides, this paper will help the researchers to better understand the CNN models for a variety of image classification task. Moreover, this paper provides a brief introduction to CNN, their applications in image processing, and discuss recent advances in region-based CNN for the past few years.
Image classification based deep learning: A Review
Aswan University Journal of Sciences and Technology
The image classification is a classical problem of image processing, computer vision and machine learning fields. Image classification is a complex procedure which relies on different components. In this paper we study the image classification using deep learning. computer vision science, image classification implementation, and deep neural networks are presented. This article discusses The development of a Convolutional Neural Network (CNN) and its various architectures, which have shown great efficiency and evaluation in image classification. A literature review is conducted to illustrate the significance and the details of Convolutional Neural Networks in various applications.
Advancements in Image Classification using Convolutional Neural Network
2018 Fourth International Conference on Research in Computational Intelligence and Communication Networks (ICRCICN), 2018
Convolutional Neural Network (CNN) is the stateof-the-art for image classification task. Here we have briefly discussed different components of CNN. In this paper, We have explained different CNN architectures for image classification. Through this paper, we have shown advancements in CNN from LeNet-5 to latest SENet model. We have discussed the model description and training details of each model. We have also drawn a comparison among those models.
International Journal of Advanced Computer Science and Applications, 2021
Face recognition is of pronounced significance to real-world applications such as video surveillance systems, human computing interaction, and security systems. This biometric authenticating system encompasses rich real human face characteristics. As such, it has been one of the important research topics in computer vision. Face recognition systems based on deep learning approaches suffer from internal covariate shift problems that cause gradients to explode or gradient disappearance, which leads to improper network training. Improper network training causes network overfitting and computational load. This reduces recognition accuracy and slows down network speed. This paper proposes a modified preactivation batch normalization convolutional neural network by adding a batch normalization layer after each convolutional layer within each of the four convolutional units of the proposed model. The performance of the proposed model is validated with a new dataset, AS-Darmaset, which is built out of two publicly available databases. This paper compared the convergence behavior of four different CNN models: the Pre-activation Batch Normalization CNN model, the Traditional CNN without Batch Normalization, the Post-Activation Batch Normalization CNN model, and the Sparse Batch Normalization CNN Architecture. The evaluation results show that the recognition performance of Pre-activation BN CNN has training and validation accuracies of 100.00% and 99.87%, the Post activation Batch normalization has 100.00% and 99.81%, and the traditional CNN without BN has 96.50% and 98.93%. The sparse batch normalization CNN has 96.25% and 97.60% success rate, respectively. The result shows that the Pre-activation BN CNN model is more effective than the other three deep learning models.
Comparative Analysis of different Convolutional Neural Network Algorithm for Image Classification
International Journal for Research in Applied Science and Engineering Technology IJRASET, 2020
In today's fast and furious world where new techniques and models are being developed every day for the sake of increasing the efficiency and performance of the neural network, this research conducts an in-depth study about some of the most popular and important Convolutional neural network models. This paper contains a detailed study and data-rich analysis of the 5 most popular Convolutional Neural networks (CNNs) for Image Detection and Identification. These 5 CNNs are LeNet, AlexNet, VGGNet16, ResNet50, and GoogLeNet. The performance of these neural networks is evaluated and benchmarked using well known and most commonly used Ciphar10 and Ciphar100 datasets. This research aims to make the process of understanding different neural networks and working with them easy. With rigorous training on the high end and graphics enabled machine for several months continuously the data and information gathered have been compiled in this research paper with all the obligatory information required to comprehend Convolutional Neural Networks.
Classification of Image using Convolutional Neural Networks
IJRASET, 2021
Now a day, with the rapid advancement in the digital contents identification, auto classification of the images is most challenging job in the computer field. Programmed comprehension and breaking down of pictures by framework is troublesome when contrasted with human visions. A Several research have been done to defeat issue in existing classification system,, yet the yield was limited distinctly to low even out picture natives. Nonetheless, those approach need with exact order of pictures. This system uses deep learning algorithm concept to achieve the desired results in this area like computer. Our framework presents Convolutional Neural Network (CNN), a machine learning algorithm is used for automatic classification the images. This system uses the Digit of MNIST data set as a bench mark for classification of gray-scale images. The gray-scale images are used for training which requires more computational power for classification of those images. Using CNN network the result is near about 98% accuracy. Our model accomplishes the high precision in grouping of images.
Journal of University of Duhok (Pure and Eng. Science), 2019
A new deep Convolutional Neural Network (CNN) with six convolutional layers and one fully-connected layer is developed and trained by backpropagation using a new optimization algorithm called Fast-AMSgrad which is modified from AMSgrad. The aims are to speed up the training process while achieving acceptable accuracy. The application of the network using both, the Fast-AMSgrad and the AMSgrad algorithms to CIFAR-10 dataset for image classification reveals that the developed CNN performs better when trained with Fast-AMSgrad for both cases, with and without Batch Normalization (BN) layers. The training time is reduced by 50% when Fast-AMSgrad algorithm is used. Also the accuracy and loss values of the training and validation are improved when Fast-AMSgrad is used. The training and validation accuracies provided by Fast-AMSgrad with BN are (91.18% and 86.92%) at epoch number (50) and (94.13% and 86.758%) at epoch number (100), while the corresponding accuracies that are provided by AMSgrad with BN are (82.65% and 81.4%) at epoch (50) and (88.82% and 85.85%) at epoch (100). The overall test accuracy and classification metric measures indicate that the given architecture of CNN and optimization algorithm perform reasonably well.
Image Classification Using Convolutional Neural Networks
International Journal of Scientific and Engineering Research, 2014
Deep Learning has emerged as a new area in machine learning and is applied to a number of signal and image applications.The main purpose of the work presented in this paper, is to apply the concept of a Deep Learning algorithm namely, Convolutional neural networks (CNN) in image classification. The algorithm is tested on various standard datasets, like remot e sensing data of aerial images (UC Merced Land Use Dataset) and scene images from SUN database. The performance of the algorithm is evaluated based on the quality metric known as Mean Squared Error (MSE) and classification accuracy. The graphical representation of the experimental results is given on the basis of MSE against the number of training epochs. The experimental result analysis based on the quality metrics and the graphical representation proves that the algorithm (CNN) gives fairly good classification accuracy for all the tested datasets.