-
翻了下好像都说的不是很详细,可否请作者详细说明下训练开源的这个模型所用的数据呢?
比如,
1.总的数据量级(有issue好像说500w左右是吗?);
2.全部都是从语料库中截取一句话来生成吗?(这样很难保证所有字符均匀出现吧?);还是随机拼接字符(这样就不好训练lstm了)?还是说按一定比例?
3.训练用了多少种字体呢?比如楷体宋体这些,测试的时候似乎都可以识别。
谢谢!
-
when I use the "crnn_main.py" to fine tune "ocr.pth" model, there is an error list:
![default](https://user-images.githubusercontent.com/22487109/46000049-1440e180-c0da-11e8-8aa1-2d83f5b91593.png)
h…
-
Is the OCR evaluation model (ster\crnn) and tatt end-to-end ? OR first use SR model to output results, and then input OCR?
just like the code below:
def __getitem__(self, index):
...
.…
-
直接拷的官网的文字识别的chinese_ocr_db_crnn_mobile模型的实例代码
![微信图片_20201207165031](https://user-images.githubusercontent.com/16176281/101329449-5ebe6b00-38ac-11eb-92c7-75d7039a4f65.png)
![微信图片_20201207165035](h…
-
~/src/ocr-lite-ncnn-1.6.1 $ ~/src/ocr-lite-ncnn-1.6.1 $ ~/src/ocr-lite-ncnn-1.6.…
-
I have done some work on OCR where I use ctpn for text detection and crnn for text recognition trained on synth90K dataset. Thz model achieves both in single approach which made me really surprised. …
-
(venv) gml@gml-TM1701:~/Documents/CHINESE-OCR$ python demo.py
Using TensorFlow backend.
Traceback (most recent call last):
File "demo.py", line 8, in
import model
File "/home/gml/Docume…
-
ImportError: No module named crnn.models.crnn,请问大神,这个怎么解决
-
1)PaddleHub和PaddlePaddle版本:PaddleHub=2.0.4,PaddlePaddle=2.1.0
2)系统环境:Linux,python版本3.7
参考https://www.paddlepaddle.org.cn/hubdetail?name=chinese_ocr_db_crnn_server&en_category=TextRecognition 部署ocr时,…
-
由于作者的代码train和validation是在一起的,我把validation的部分提出来单独用作者训练好的模型对360万数据集的36万多张图片做validation,但是效果并不好,val accuracy只有78.4%
![image](https://user-images.githubusercontent.com/48081275/63398740-1520e700-c401-11…