Abstract
As a part of image recognition, Chinese character recognition has a great application market in China, such as license plate recognition, logistics information recognition and so on. In recent years, the application of CNN has set off a frenzy of computer vision. Especially in the task of image recognition, CNN is widely used because of its high accuracy and few calculation parameters. However, CNNs local operation of using repeated filters to process images also has its shortcomings, for example, it cant pay attention to the relationship between distant pixels in the image. To solve this problem, we tried to add non-local operation to CNN to improve its performance. We chose the classic model of ResNetV2-50 as the foundation, and added non-local blocks to it. We compared the results of the two models and found that the accuracy increased by 4%.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
