search

liuheng92 / tensorflow_PSENet

This is a tensorflow re-implementation of PSENet: Shape Robust Text Detection with Progressive Scale Expansion Network.My blog:
https://blog.csdn.net/liuxiaoheng1992/article/details/87646951
MIT License
489 stars 162 forks source link

Training loss not changing #87

Open uraibeef opened 4 years ago

uraibeef commented 4 years ago

Hello, i try to re-train on my dataset in the same format like your training dataset, and when i training doesn't have an error but my model training loss not changing it's stay at 1.0303 every step.. What's wrong? and what should i do?

wdh234 commented 3 years ago

训练数据集是否有模糊标注，即1477,609,1556,604,1558,633,1478,638,####

lipeng1109 commented 3 years ago

我也有同样的问题，请求帮助

uraibeef commented 3 years ago

训练数据集是否有模糊标注，即1477,609,1556,604,1558,633,1478,638,####

ในวันที่ อ. 2 ก.พ. 2021 เวลา 10:02 lipeng1109 notifications@github.com เขียนว่า:

我也有同样的问题，请求帮助

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/liuheng92/tensorflow_PSENet/issues/87#issuecomment-771316645, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJMBWTKSYXWKDIDAQMXYXQ3S45TNFANCNFSM4QJKHQBA .