This is KOREAN version code of original SynthText scene-text image generator. (You can find the English, original version in here: https://github.com/ankush-me/SynthText)
Code for generating synthetic text images as described in "Synthetic Data for Text Localisation in Natural Images", Ankush Gupta, Andrea Vedaldi, Andrew Zisserman, CVPR 2016.
Korean Synthetic Scene-Text Image Samples
The main dependencies are:
pygame, opencv (cv2), PIL (Image), numpy, matplotlib, h5py, scipy
The 8,000 background images used in the paper, along with their segmentation and depth masks, have been uploaded here:
http://zeus.robots.ox.ac.uk/textspot/static/db/<filename>
, where, <filename>
can be:
imnames.cp
[180K]: names of filtered files, i.e., those files which do not contain textbg_img.tar.gz
[8.9G]: compressed image files (more than 8000, so only use the filtered ones in imnames.cp)depth.h5
[15G]: depth mapsseg.h5
[6.9G]: segmentation maps
Note: I do not own the copyright to these images.
The background image files are saved in bg_img
directory. The following files are used to generate the images.
- data/fonts: several sample fonts (add more fonts to this folder and then update
fonts/fontlist.txt
with their paths). You must put Korean fonts, otherwise you cannot print out the letters in the picture properly. - data/newsgroup: Text-source (from the News Group dataset). This can be subsituted with any text file. Look inside
text_utils.py
to see how the text inside this file is used by the renderer. - data/models/colors_new.cp: Color-model (foreground/background text color model), learnt from the IIIT-5K word dataset.
- data/models: Other cPickle files (char_freq.cp: frequency of each character in the text dataset; font_px2pt.cp: conversion from pt to px for various fonts: If you add a new font, make sure that the corresponding model is present in this file, if not you can add it by adapting
invert_font_size.py
).
python make_h5.py
This will generate random scene-text image files with background images and store them in an h5 file in dset_kr.h5
. When it generates the image, it also creates the following files.
word_text.txt
: words that are generated from scene-text.char_text.txt
: characters that are splited from words.
python visualize_results.py
This will generate the followings from stored images in dset_kr.h5
. There are 3 types of dataset that you can generate.
~/images
: Scene-text image files~/words
: Word image files from~/images
~/characters
: Character image files from~/images
Also, you can get the coordinate files of characters and words each
ch_coords.txt
: Characters' each coordinate in scene-text image fileswd_coords.txt
: Words' each coordinate in scene-text image files
Segmentation and depth-maps are required to use new images as background. Sample scripts for obtaining these are available here.
predict_depth.m
MATLAB script to regress a depth mask for a given RGB image; uses the network of Liu etal. However, more recent works (e.g., this) might give better results.run_ucm.m
andfloodFill.py
for getting segmentation masks using gPb-UCM.
For an explanation of the fields in dset.h5
(e.g.: seg
,area
,label
), please check this comment.
Please refer to the paper for more information, or contact the author (email address in the paper). If you find errors for koreans, don't hesitate to give me an email: [email protected]