this project is very helpful for generating synth text for scene text recognition, and it seem to generate text image by combine several character images, but the outputs doesn't contain information of each character, is it possible to get character annotation, for example, each character and its location?

Hi, @moonbings Thank you for sharing nice work.

I wonder where's the code that load background image such as below?

And I also wonder when the Synthdog will be released

Hi, thank you for open-sourcing your work. I have a question about the training configuration of STR model used in your paper. How did you set the sensitive character mode and data_filtering_off option in BEST model?

On running gen.py with the command posted in the readme, the process throws

munmap_chunk(): invalid pointer

and is stuck. It stops generating images. Each worker generates a maximum of 1 image before this error is thrown.

Thanks for the nice package! When i followed your descriptions and used it, i met some error. KeyError: 'setting text direction, language or font features is not supported without libraqm' I have installed python package 'synthtiger' and dependency following the shell script How can i solve this issue? Thanks,

thank you for your awesome sharing. i tried to generate text recognition data using your library for bangla printed documents,i made all the necessary changes to make this synthesis engine working for bengali text recognition data generation, however for bangla it is breaking text during text to image conversion just like this trdg issue of mine : https://github.com/Belval/TextRecognitionDataGenerator/issues/253 (where the authors of that repo is not responsive at all)

here is a sample that i got while using this synthesis engine.

label : নির্লজ্জ

image : https://i.ibb.co/0Y0W5GG/image.jpg ( নির্লজ্জ word got broken )

do you know how to solve this issue? thanks a lot in advance for your great work.

Hi, great work ! What exactly should one modify to change language. I found this: https://github.com/clovaai/synthtiger/tree/master/resources/charset but not sure if I need to generate something from that.

thanks in advance :)

Description

• Fixed a bug for seed with none value. (https://github.com/clovaai/synthtiger/pull/45#issuecomment-1311229534)
• Added retry parameter in generator function to make testing easier.
  • If retry is False, when an error occurs during data generation, it does not retry and return data of None value.
• Updated test code.

Changes in this PR

• Updated main.py, gen.py.
• Updated test code.

How has this been tested?

Checked directly whether images are generated correctly.

Description

• Random seed is also applied to main process.
• Added get_global_random_states, set_global_random_states, set_global_random_seed functions. This functions can be used in templates.

Changes in this PR

• Updated main.py, gen.py.
• Added functions (get_global_random_states, set_global_random_states, set_global_random_seed).

How has this been tested?

Checked directly whether images are generated correctly.

Description

This engine cannot reproduce same data. So I added random seed option to solve this issue.

I also changed multi processing structure. When the new engine puts tasks in the task queue, each worker gets and runs it. And each task is assigned an index. Note that the incoming indexes are random when storing data because the task completion times are different.

Changes in this PR

• Added random seed option. (-s or --seed)
• Refactored main.py, gen.py
• Updated README.md
• Updated .pylintrc

How has this been tested?

Checked directly whether images are generated correctly.

Description

Updated shell script to get latest package list before installing dependencies in ubuntu.

Changes in this PR

• Updated install_ubuntu_depends.sh.

How has this been tested?

Check CI test.

Hi All,

I have downloaded all parts of synthtiger_v1.1.zip. When I combined into a single zip file, the resulting zip file is not valid.

Anyone faces similar problem ?

Appreciate your advices.

Hi team,

Is there any chance you could improve the documentation for this library to describe what the functions and classes actually do? I'm having trouble figuring out how to utilize this library for different kinds of unstructured & semi-structured text images.

Hello @moonbings During synthetic dataset generation, some image are very distorted and i don't have any idea how to fix it. I played with some parameters but it didnt work for me. Any solution?

Samples image are not clear

Hello @moonbings While generating Korean dataset, I have added all given requirements like fonts and everything. I am using below command to generate dataset but getting an error between data generation process.

python -m synthtiger -o results -w 4 -v examples/synthtiger/template.py SynthTiger examples/synthtiger/config_vertical.yaml

Generated 56 data
Generated 57 data
Generated 58 data
Generated 59 data
Generated 60 data
Traceback (most recent call last):
  File "/media/cvpr/CM_22/synthtiger/synthtiger/gen.py", line 71, in _generate
    data = template.generate()
  File "/media/cvpr/CM_22/synthtiger/examples/synthtiger/template.py", line 117, in generate
    image = _blend_images(fg_image, bg_image, self.visibility_check)
  File "/media/cvpr/CM_22/synthtiger/examples/synthtiger/template.py", line 246, in _blend_images
    raise RuntimeError("Text is not visible")
RuntimeError: Text is not visible

Generated 61 data
Generated 62 data
Generated 63 data
Generated 64 data
Generated 65 data
Generated 66 data