-
-
Notifications
You must be signed in to change notification settings - Fork 539
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
关于识别电子称读数的问题。 #292
Comments
群二维码过期了,望更新一下。 |
请问你的数据集是不定长的图片和标签吗,你是如何修改代码的呢 |
您好,我也是训练数码管数字,刚开始预测有输出,后来全是blank,请问怎么解决这个问题呢 |
我在训练电子称读数识别时,验证集部分效果如下:
0000..-----------------------------00--00 => 0.00 , gt: 0.00
9999----------..----------------------511 => 9.51 , gt: 90.51
33-6-------------.-------------------0022 => 36.02 , gt: 36.02
116-------------------------------------- => 16 , gt: 16.87
2277------------..--------------------888 => 27.8 , gt: 27.88
11222--------------.-----------------9933 => 12.93 , gt: 12.93
11144-----------------..--------------555 => 14.5 , gt: 134.25
0000-..----------------------------00--00 => 0.00 , gt: 0.00
Test loss: 0.3058, accuray: 0.6571
然后运行demo.py,正确率为0%,部分结果如下:
tag:36.02 pred:1012
tag:62.25 pred:12
tag:0.00 pred:10
tag:14.51 pred:151
tag:65.65 pred:2
tag:0.00 pred:100
我觉得可能是预测的标签太多(预测标签数41),而字符串长度最长才6,或许应该修改预测标签数,但是我查看了代码,不知道在哪里修改,
也有可能是数据太少(train:1128张图片, val:280张图片),字典为'0123456789.',应该怎么优化可以提高模型效果,望告知。
The text was updated successfully, but these errors were encountered: