The official implementation of the "A high-performance CNN method for offline handwritten Chinese character recognition and visualization" paper, Soft Comput 24, 7977–7987 (2020). https://doi.org/10.1007/s00500-019-04083-3
Hello, Why I cannot get the accuracy proposed in the paper, although I performed the experiment according to the guides same as that of paper? Please give me some instructions if possible?Thanks a lot!
Hello, my name is Yun Rui Li. I am a graduate student of the National Tsing Hua University Department of Computer Science.
Our laboratory( http://make.cs.nthu.edu.tw/makeWeb/ )is doing research on ASD, and my work is to collect the Chinese characters written by ASD and TD, evaluate the aesthetics of the characters, analyze the scoring model, and hope to find out the characteristics of ASD.
I am interested in the studies mentioned in “A high-performance CNN method for offline handwritten Chinese character recognition and visualization”. Because I think the HCCR model is helpful for aesthetic evaluation, and the CAM part is good for analyzing the model. I hope to fine-tune this model and use it in my work, but after I ran the code you released, I found that only the prediction part.
(I have previously sent letters to the three authors of the paper and received a reply from Zhiqiang You, he said that the code is maintained by the first author)
Is it convenient for you to provide the code of the training part?
Thank you very much for your help.
Hey there thank you for this repository, I was wondering if it's possible to use your code to train the network for different dataset (like in a different language character recognition task)? If that is possible then which code and also which part of the code should I be looking for could you please specify? Thank you.