fakeryfx / incepttext-tensorflow

An Implementation of the alogrithm in paper IncepText: A New Inception-Text Module with Deformable PSROI Pooling for Multi-Oriented Scene Text Detection

Introduction

• email:[email protected]
• 微信公众号《计算广告那些事儿》话事人
• 个人博客:CSDN
• 🎓 Master of Computer Science
• 🔭 Search/Ads/Rec

Visitor count

incepttext-tensorflow's Issues

about detection performance

how about the result on ICDAR15 ?

make lib wrong

the info is bellow

4 errors detected in the compilation of "/tmp/tmpxft_0000762e_00000000-6_deform_psroi_pooling_op_gpu.cu.cpp1.ii".

could you help me?

关于deform_conv.so

cd /lib;./make.sh error

/python2.7/site-packages/tensorflow/include/unsupported/Eigen/CXX11/../../../Eigen/src/Core/MathFunctions.h(1265): warning: calling a constexpr host function("real") from a host device function("abs") is not allowed. The experimental flag '--expt-relaxed-constexpr' can be used to allow this
lib/python2.7/site-packages/tensorflow/include/unsupported/Eigen/CXX11/../../../Eigen/src/Core/MathFunctions.h(1270): warning: calling a constexpr host function("real") from a host device function("abs") is not allowed. The experimental flag '--expt-relaxed-constexpr' can be used to allow this.
why?

the test.py have something wrong?

$python test.py
File "test.py", line 319
t_00 = int(box[0,0])
^
TabError: inconsistent use of tabs and spaces in indentation

python test.py No module named 'lanms'/'locality_aware_nms'

Could you give the test demo?

thank you for releasing the incepttext code.

Could you share the test demo?

thank you very much

wrong:poly in wrong direction....when i trian on the data

invalid poly
invalid poly
Step 000220, model loss 0.0331, total loss 0.3090, 2.03 seconds/step, 3.44 examples/second
Step 000230, model loss 0.0496, total loss 0.3252, 2.03 seconds/step, 3.44 examples/second
poly in wrong direction
poly in wrong direction
Step 000240, model loss 0.0279, total loss 0.3031, 2.13 seconds/step, 3.28 examples/second
poly in wrong direction
poly in wrong direction
Step 000250, model loss 0.0298, total loss 0.3047, 2.25 seconds/step, 3.10 examples/second

The model is based on EAST, NOT the IncepText. The loss is tend to nan after more than 20000 iters. How to solve this problem?

The model is based on EAST, NOT the IncepText. The loss is tend to nan after more than 20000 iters. How to solve this problem?
Can you post your results in ICDAR 2015/2017?

模型中Deformable PSROI Pooling代码去掉了吗？

作者您好，Deformable PSROI Pooling部分的代码没有找到，能指点一下在哪吗，还是已经去掉了

Could share the pre-trrianed model?

Could share the pre-trrianed model?

thank you very much.

Could you share your trained modle which you tained on the data?

Thank you for contributing to this article.

I want to test some icdar data imge on the trained model.

Could share the model ?

thank you very much

Look forward for your reply.

Loss diverged, stop training

t1 is always nan。。。

Documentation on Training

@xieyufei1993 Thank you for your hard work
Can you share more documentation on training

deform_psroi_pooling_op.cu.o: No such file or directory

@xieyufei1993 Do you know how to solve this problem ?I was confused.
Thank you very much.