Laser & Optoelectronics Progress, Volume. 57, Issue 8, 081002(2020)
Handwritten Chinese Character Recognition Based on Attention Mechanism
Figures & Tables(7)
Table 1. Structure and parameters of the AT-CNN model
View tableTable 1. Structure and parameters of the AT-CNN model
Sequence of layer Feature map size Number of feature maps Filter size Step size C1 64×64 32 3×3 1 MP2 32×32 32 2×2 2 C3 32×32 64 3×3 1 MP4 16×16 64 2×2 2 C5 16×16 128 3×3 1 MP6 8×8 128 2×2 2 C7 8×8 256 3×3 1 C8 8×8 256 3×3 1 MP9 4×4 256 2×2 2 FC 1×1 1024 — — Output 1×1 3755 — —
Table 2. Comparison of the recognition accuracy of different methods on HWDB dataset
View tableTable 2. Comparison of the recognition accuracy of different methods on HWDB dataset
Recognition method RAM/MB Accuracy /% MQDF[ 20 ]— 89.55 CCPR-2010 champion: HKU[ 21 ]339. 10 89. 99 MQDF+CNN[ 20 ]— 92.03 ICDAR-2011 champion: IDSIAnn-2[ 22 ]71. 35 92. 18 ICDAR-2013 champion: Fujitsu[ 3 ]2460. 00 94. 77 HCCR-Ensemble-GoogLeNet[ 11 ]277. 25 96. 74 CNN 108.30 93.01 AT-CNN 115.30 95.05
Tools
Get Citation
Copy Citation Text
Wanrong Huang, Kai He, Kun Liu, Shengnan Gao. Handwritten Chinese Character Recognition Based on Attention Mechanism[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081002
Paper Information
Category: Image Processing
Received: Jul. 5, 2019
Accepted: Sep. 10, 2019
Published Online: Apr. 3, 2020
The Author Email: Kai He (hekai@tju.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20