Can I use this tool to do scene text detection + recognition ?
lomograb opened this issue · 2 comments
lomograb commented
I wonder if I could train the network with pieces of grayscaled image from camera instead of black-in-white document only. Do LSTM has coverage about this case ?
wanghaisheng commented
maybe this one would be helpful
lomograb commented
I just discover that crnn is incredibly slow. It took about 3 seconds per pieces of an image to recognize all the texts. So if I have 10 text boxes in an image, it would take 30~ seconds to finish