A Qualitative Analysis of Googlenet and Alexnet for Fabric Defect Detection
K.K. Sudha1, P. Sujatha2
1K. K. Sudha, Research Scholar, Department of Information Technology, Vels Institute of Science, Technology & Advanced Studies (VISTAS), Chennai, India,
2Dr. P. Sujatha, Professor, Department of Information Technology, Vels Institute of Science, Technology & Advanced Studies (VISTAS), Chennai, India.
Manuscript received on 20 April 2019 | Revised Manuscript received on 05 May 2019 | Manuscript published on 30 May 2019 | PP: 86-92 | Volume-8 Issue-1, May 2019 | Retrieval Number: A2959058119/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: In this paper, the performance of Convolutional Neural Networks such as GoogleNet (Inception) and AlexNet are analyzed for the textile defect detection problem. The fabric images from ‘Cotton Incorporated’ database is used for this research work. The database images are converted to grey scale. The noises are removed from the grey scale image using Wiener filter. The noise free images are trained using GoogleNet and AlexNet to recognize new faults in the fabric. The identification of fabric fault by using GoogLeNet include image load, loading GoogLeNet network, loading pretrain network, freezing of the basic layers, and image validation. The steps in AlexNet for finding the fabric defects are image load, AlexNet network load, substitution of the final layers, training network, and image classification. According to the results of the experiment, Goog LeNet training on fabric defects is faster than that of Alex Net. The performance of GoogLeNet is the best outdoing than Alex Net on various parameter including time, accuracy, dropout, and the initial learning.
Index Terms: Google Net, Alex Net, Convolutional Neural Network, Scale Range, Dropout Ratio.
Scope of the Article: Bio – Science and Bio – Technology