Algorithm of Scene Text Detection

####1.Description:
this program show a demo of text detection using the Algorithm in
SCENE TEXT DETECTION WITH EXTREMAL REGION BASED CASCADED FILTERING

####2.Needed Materials (important):
(1)Models
Models are on : 3rd_party_tools\
broad_pca_data_Dec_06 pca_data of broad type
narrow_pca_data_Dec_06 pca_data of narrow type
broad_svmStruct_Dec_06 broad SVM
narrow_svmStruct_Dec_06 narrow SVM
detnet_layers cnn model

(2) Matlab platform is needed. the code tests on Matlab 2014b.

(3) You need a windows7 x64 system with CUDA 6.5 or a linux system with CUDA7.0.
if you have other version of CUDA on either system, you need to compile the Matconvnet and get your own mex file, because the off-the-shelf model in [1] is used, for more information, see the use of model in Deep features for text spotting

(3) data
you can use img_1.jpg available or your own data , you need to change the path in single_img_pipeline.m

####3. Use
run single_img_pipeline.m

[1] M.Jaderberg, A.Vedaldi, and A.Zisserman, “Deep features for text spotting” in Proceeding of ECCV, 2014, pp. 512–528.