The Study and Analysis of Different CNC Model Based on Image CAPTCHA Recognition

DOI

10.2991/aer.k.201203.032How to use a DOI?

Keywords

TensorFlow, CAPTCHA, CNN, Accuracy, Model Training

Abstract

With the development of the AI technology, Machine learning and deep learning have been paid attention on increasing extensively. The neural network in deep learning, especially the CNN (Convolutional Neural Network) were used in aspect of image processing and many kinds of AI projects. Deep learning is good at image processing which has resulted in outstanding performance. At the same time, people have difficulties in choosing different CNN model when they use it. This paper picked up a classic project CAPTCHA as a sample. We designed different kinds of CNN model after the studding of LeNet-5 network and VGG GoogleNet etc. First of all, we study the result and the accuracy percent with the same number of CNN layers in different strides, different kernel size and different maps. Then with the rise of the number of hidden CNN layers, two more CNN model were studied with same parameters. We try to find out the relationship between the accuracy percentage and different CNN models. We hope it can do any help to the people who use the CNN model to do some AI project. A kind of the best models was fond in the progress of our studying, it was proved having a good performance not only in the validate accuracy but also in the test accuracy.

Copyright

© 2020, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Ligong Cui
AU  - Yanyan Cao
AU  - Liqiang Nie
PY  - 2020
DA  - 2020/12/03
TI  - The Study and Analysis of Different CNC Model Based on Image CAPTCHA Recognition
BT  - Proceedings of the 2020 9th International Conference on Applied Science, Engineering and Technology (ICASET 2020)
PB  - Atlantis Press
SP  - 170
EP  - 175
SN  - 2352-5401
UR  - https://doi.org/10.2991/aer.k.201203.032
DO  - 10.2991/aer.k.201203.032
ID  - Cui2020
ER  -