Codes and transferred deep model of my ICTAI 2017 paper "Using transferred deep model in combination with prior features to localize multi-style ship license numbers in nature scenes", please refer our paper for more details.
- The Transferred_Deep_Model was implemented and trained with Caffe (TextBoxes version of Caffe, see also https://github.com/MhLiao/TextBoxes).
- How to use ? To combine our transferred deep model and other two SLNs prior-based SLNs region generating and fake SLNs filtering algorithms (they are implemented with C++), we use an Eclipse IDE for C/C++ Developers. By calling our transferred deep model and two algorithms seamlessly, we can conduct an end-to-end style testin on this Eclipse platform.
You are advised to figure out how to use C++ to handle Caffe-based projects.You may find a sample code in "~/caffe/examples/ssd/ssd_detect.cpp" from SSD (https://github.com/weiliu89/caffe/tree/ssd). Once you figure this point out, it will be easy for you to run our codes and transferred deep model with an Eclipse IDE for C/C++.
Please let me know if you encounter any issues.
P.S: Our transferred deep model is too big to be uploaded on Github, you can find it at http://pan.baidu.com/s/1i4YJHLb