Research Article
Performance Analysis of Deep Neural Networks Using Computer Vision
@ARTICLE{10.4108/eai.13-10-2021.171318, author={Nidhi Sindhwani and Rohit Anand and Meivel S. and Rati Shukla and Mahendra Pratap Yadav and Vikash Yadav}, title={Performance Analysis of Deep Neural Networks Using Computer Vision}, journal={EAI Endorsed Transactions on Industrial Networks and Intelligent Systems}, volume={8}, number={29}, publisher={EAI}, journal_a={INIS}, year={2021}, month={10}, keywords={Computer Vision, Object Detection, Deep Learning, Deep Neural Network, Image classification, Artificial intelligence, Machine learning, Convolutional neural network, Accuracy}, doi={10.4108/eai.13-10-2021.171318} }- Nidhi Sindhwani
Rohit Anand
Meivel S.
Rati Shukla
Mahendra Pratap Yadav
Vikash Yadav
Year: 2021
Performance Analysis of Deep Neural Networks Using Computer Vision
INIS
EAI
DOI: 10.4108/eai.13-10-2021.171318
Abstract
INTRODUCTION: In recent years, deep learning techniques have been made to outperform the earlier state-of-the-art machine learning techniques in many areas, with one of the most notable cases being computer vision. Deep learning is also employed to train the neural networks with the images and to perform the various tasks such as classification and segmentation using several different models. The size and depth of current deep learning models have increased to solve
certain tasks as these models provide better accuracy. As pre-trained weights may be used for further training and prevent costly computing, transfer learning is therefore of vital importance. A brief account is given of their history, structure, benefits, and drawbacks, followed by a description of their applications in the different tasks of computer vision, such as object detection, face recognition etc.
OBJECTIVE: The purpose of this paper is to train a deep neural network to properly classify the images that it has never seen before, define techniques to enhance the efficiency of deep learning and deploy deep neural networks in various applications.
METHOD: The proposed approach represents that after the reading of images, 256x256 pixel image’s random parts are extracted and noise, distortion, flip, or rotation transforms are applied. Multiple convolution and pooling steps are applied by controlling the stride lengths.
RESULT: Data analysis and research findings showed that DNN models have been implemented in three main configurations of deep learning: CNTK, MXNet and TensorFlow. The proposed work outperforms the previous techniques in predicting the dependent variables, learning rate, image count, image mean, performance analysis of loss rate and learning rate during training, performance Analysis of Loss with respect to Epoch for Training, Validation and Accuracy.
CONCLUSION: This research encompasses a large variety of computer applications, from image recognition and machine translation to enhanced learning. DNN models have been implemented in three main configurations of deep learning: CNTK, MXNet and TensorFlow. Extensive research has been conducted using the various deep architectures such as AlexNet, InceptionNet, etc. To the best of authors’ knowledge, this is the first work that presents a quantitative analysis of the deep architectures mentioned above.
Copyright © 2021 Nidhi Sindhwani et al., licensed to EAI. This is an open access article distributed under the terms of the Creative Commons Attribution license, which permits unlimited use, distribution and reproduction in any medium so long as the original work is properly cited.