some questions about char2wordBB and the generated text

[Muran337287]

1. I don't understand the function char2wordBB, what's the meaning of "cc = np.squeeze(np.concatenate(np.dsplit(cc,cc.shape[-1]),axis=1)).T.astype('float32')" and " cc_tblr"? Someone can help me?

2. I met some cases with the generated text.

1. The text is underlined. "know" in the first pic.
2. "." and text overlap. "rec.sport.hockey" in the second pic.
3. the bounding box may appear to be not tightly enclosed. "27" and “Apr”in the third pic.
4. word is not rendered successfully. "files" and "that" in the fourth pic.

[yujunsun110]

"Files" and "that" look good to me, the word color is very close to the background color.

[chaitusvk]

Which version Are you using python2 or python3 model ....