Official code: https://github.com/MhLiao/TextBoxes_plusplus Paper: https://arxiv.org/abs/1801.02765
we pre-train TextBoxes++ on SynthText dataset for all tested datasets. Then the training process is continued on the corresponding training images of each dataset.