** Build an ID Card recognition API with implementing different image processing methods on top of Text-Detection-CTPN (below) to better tackle text detection in old, damaged and artifacts-deposited ID cards **
Scene text detection based on ctpn (connectionist text proposal network). It is implemented in tensorflow. The origin paper can be found here. Also, the origin repo in caffe can be found in here. For more detail about the paper and code, see this blog.
nms and bbox utils are written in cython, hence you have to build the library first.
cd utils/bbox chmod +x make.sh ./make.sh It will generate a nms.so and a bbox.so in current folder.
follow setup to build the library download the ckpt file from google drive or baidu yun put checkpoints_mlt/ in text-detection-ctpn/ put your images in data/demo, the results will be saved in data/res, and run demo in the root python3 ./main/demo_al.py