Text Detection in images using mxnet in R


#1

Hi,

This is the first time I am attempting to use mxnet. However, the problem at hand is a bit challenging and I am unable to get any good examples. I need to create a model that can detect text data in images. I have tried the tesseract package in R. But the results are not very good. Any help in this regard would be great.


#2

Hi FinScience,
Unfortunately we do not have a tutorial on OCR yet in the Apache MXNet repo.
We have an example here: using LSTM+CTC:


You can find here some code using MXNet performing OCR in the wild:

Based on this paper:
https://arxiv.org/abs/1707.08831
Hope that helps,


#3

I have actually written a notebook to do OCR on text using CNN + biLSTM + CTC using Gluon. You can check it out here: https://github.com/ThomasDelteil/Gluon_OCR_LSTM_CTC It still need some polishing to be an official tutorial, but that should get you started.