text-detection-ctpn-pytorch
my blog about CTPN blog
setup
nms and bbox utils are written in cython, you have to build the library first.
cd utils/bbox
sh make.shIt will generate a nms.so and a bbox.so in current folder.
how to test
- follow setup to build the library
- download the test model from baidu yun
- change you own model_path , dir_path and save_path in inference.py
python3 inference.py
how to train
####data format follow icdar15 dataset format, x1,y1,x2,y2,x3,y3,x4,y4,label
image
│ 1.jpg
│ 2.jpg
│ ...
label
│ 1.txt
│ 2.txt
| ...
get pre_model
- First, download the pre-trained model from torchvision.models.vgg, put it in root_dir .
python3 get_premodel.py
it will generate the pre_model named "vgg16.model" in root_dir
then change you train data path in ./dataloader/data_provider.py , the train data path is the root path of your image and label dir
train
Simplely run
python3 train.py
some results
#reference