Acta Photonica Sinica, Volume. 51, Issue 11, 1111001(2022)
Handwritten Font Classification Method Based on Ghost Imaging
Figures & Tables(7)
Fig. 1. The realization process of automatic recognition method of handwritten digit
Fig. 3. The accuracy of digit from 0 to 9 under different sampling rates of CNN and DNN models
Fig. 4. Accuracy rate of 10 letters under different sampling rates and CNN model precision，recall rate，F1 score
Fig. 5. The loss error curve of different models using the cross-entropy loss function
Table 1. Comparison of accuracy based on CNN and DNN models
View tableView in ArticleTable 1. Comparison of accuracy based on CNN and DNN models
Model Lowest accuracy Highest accuracy Average accuracy CNN 95.65% 98.47% 97.65% DNN 87.56% 92.99% 92.34%
Table 2. Comparative analysis of precision，recall rate and F1 score under DNN and CNN models
View tableView in ArticleTable 2. Comparative analysis of precision，recall rate and F1 score under DNN and CNN models
Parameter DNN CNN Lift value Precision 86.50% 97.25% 10.75% Recall rate 86.40% 98.03% 11.63% F1 score 86.31% 97.60% 11.09%
Tools
Get Citation
Copy Citation Text
Ruyu YAN, Xiaoxia WANG, Jiangtao XI, Fengbao YANG, Daerhan BAO. Handwritten Font Classification Method Based on Ghost Imaging[J]. Acta Photonica Sinica, 2022, 51(11): 1111001
Paper Information
Category:
Received: Feb. 6, 2022
Accepted: Mar. 14, 2022
Published Online: Dec. 13, 2022
The Author Email: WANG Xiaoxia (wangxiaoxia@nuc.edu.cn), XI Jiangtao (20180143@nuc.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20