-
Notifications
You must be signed in to change notification settings - Fork 28
/
Copy pathREADME.md
76 lines (42 loc) · 2.87 KB
/
README.md
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
# SRNet-Datagen - A data generator of SRNet
## Introduction
This is a data generator of SRNet which is the model described in the paper *Editing Text in the wild*.
Our tensorflow reproducing of SRNet: [https://github.com/youdao-ai/SRNet](https://github.com/youdao-ai/SRNet)
Original paper: [*Editing Text in the wild*](https://arxiv.org/abs/1908.03047) by Liang Wu, Chengquan Zhang, Jiaming Liu, Junyu Han, Jingtuo Liu, Errui Ding and Xiang Bai.
This data generator project is a simplification based on the following two projects.
[Synthtext](https://github.com/ankush-me/SynthText): Extracted the rendering part of the project and Adjusted to Python3 code to get `i_s`, `t_t`, `t_f` and `mask_t`
[Skeletonization-of-Digital-Patterns](https://github.com/anupamwadhwa/Skeletonization-of-Digital-Patterns): Adjusted this project to Python3 code to skeletonize mask_t and get `t_sk`.
## Generate data
First prepare a directory of fonts and a background datasets without text. You can also prepare a word corpus for rendering.
You need to write the absolute path of each data in the background dataset as a line into a file, and modify `bg_filepath` parameter of `Synthtext/data_cfg.py` to the path of this file.
You can adjust other data configurations in `Synthtext/data_cfg.py`. The following is a description of some parameters.
- `font_dir`: the directory path of fonts in ttf format.
- `standard_font_path`: the standard font to render i_t.
- `text_filepath`: a file containing the text of the word to be rendered, each line is a word.
- `bg_filepath`: a file containing the absolute path of each background image.
- `color_filepath`: a file used to select the color of the text which is given by Synthtext project.
Then you will need to adjust generating configurations in `cfg.py` including saving directory, the amount of data to generat and the number of processes that are needed.
Finally `python3 datagen.py` and start generating.
You can also use this project to generate data online while training SRNet.
- `i_s`: styled text a rendering on background image
- `i_t`: standard text b rendering on gray background
- `t_sk`: skeletonization of styled text b.
- `t_t`: styled text b rendering on gray background
- `t_b`: background image
- `t_f`: styled text b rendering on background image
- `mask_t`: the binary mask of styled text b
From left to right, from top to bottom are examples of `i_s, i_t, t_sk, t_t, t_b, t_f, mask_t`
## Requirements
- Python 3.6
- numpy
- opencv-python
- Augmentor
## Reference
- [Editing Text in the Wild](https://arxiv.org/abs/1908.03047)
- [Synthetic Data for Text Localisation in Natural Images](https://arxiv.org/abs/1604.06646)
- [A fast parallel algorithm for thinning digital patterns](http://www-prima.inrialpes.fr/perso/Tran/Draft/gateway.cfm.pdf)