The released trained model - what datasets was it trained on?

[ramanishka]

@MhLiao Thanks for your work!

[MhLiao]

@ramanishka It was trained by a combination of several datasets including SynthText, ICDAR 2015 dataset, ICDAR 2013 dataset, SCUT dataset, and Total-Text dataset. More details are in Sec 4.2.1 in the paper.