text-detection-train-yolov3
text-detection-train-yolov3 copied to clipboard
Convolutional Neural Networks

Darknet
Darknet is an open source neural network framework written in C and CUDA. It is fast, easy to install, and supports CPU and GPU computation.
For more information see the Darknet project website.
For questions or issues please use the Google Group.
some using detaile can be found in below link:
https://github.com/AlexeyAB/darknet
text-detection-yolov3
this project is used to detect text mainly reference:
https://github.com/Aurora11111/keras_ocr
https://github.com/tianzhi0549/CTPN
https://github.com/pjreddie/darknet.git
datasets prepare
first, you can down load icdar2017:链接:http://rrc.cvc.uab.es/?ch=7&com=downloads
then :python voc_annotation.py
if your dataset format is json:
python dict_datasets.py
attentions 1:the accordibnate sequences must be:(xmin, xmax,ymin, ymax)
generate anchors
python kmeans.py
attentions 1:the accordibnate sequences must be:(xmin, ymin, xmax, ymax)
attentions 2:file structural must be similarity to voc
training
./darknet detector train cfg/voc.data cfg/text.cfg darknet53.conv.74