1LOVESJohnny
1LOVESJohnny
change the actual data directory as you put them local.
> 识别还是有问题，现在没空弄了。 Hi, how about now?
+1 Waiting for the workaround... Or only we can change the network architecture?
> CMake build fails with missing package `FindOpenEXR.cmake`: > > ``` > CMake Error at openexrid/CMakeLists.txt:1 (FIND_PACKAGE): > By not providing "FindOpenEXR.cmake" in CMAKE_MODULE_PATH this project has > asked CMake...
同问如何生成的gt_file，另外，请问哪里可以找到原始的txt标注数据呢？ 云盘下载的数据是xml格式，但是xml里框的数量和图像上的字符数量对应不上。想请问是如何生成的？
> 训练集都是白底黑字的样本，想识别黑底白字得加入相应样本，如果你的场景只是黑底白字就把样本做个黑白反转再训练 您好，如果我的场景需要识别彩色场景里的文字，是否需要将数据集替换成彩色数据？另外，是否需要对应修改数据读入gen函数？
Hey guys, I've found a solution. source answer: https://www.jianshu.com/p/e948f44c6b37 `Error info: THC_API cudaError_t THCudaMalloc(THCState *state, void **ptr, size_t size); error: command 'x86_64-linux-gnu-gcc' failed with exit status 1. Modification: pytorch_binding/src/binding.cpp: 1...
Since the update of opencv, I've found this solution. You can replace all the 'cv.cv' into 'cv' only. And correspondingly, change CV_RGB2Lab into COLOR_RGB2Lab. Hope it could help.
> i have the same problem with you, have you solved it now? if so ,can you tell me what you have done ? replace cv.cv.CV_RGB2Lab with cv.RGB2Lab
> Maybe there are some mistakes in the annotation's format of your dataset. > > In our settings, the format of the annotations in character-level: > img: icdar2013 train set...