Giters

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.

Home Page:http://www.robots.ox.ac.uk/~vgg/data/scenetext/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Why can TextSource.sample_line return multiple lines?

GPhilo opened this issue · comments

commented

Hello,

looking at the code of TextSource, I see that sample_line's implementation randomly picks a number of lines between 1 and 3 (with weights [0.85, 0.10, 0.05]).
Why is this the case? I would assume the function to sample just one line, relying on sample_para for multiline cases (I'm aware of the optional line centering applied to paragraphs, but when that is not active "line" and "paragraph" are indistinguishable), so I don't understand the design choice.
Could the author (or someone with more familiarity with the code) clarify the design choice?

Thank you!

commented

This is research codebase (not production ready), and can certainly be improved.