2
$\begingroup$

I have been working to design a system that uses multiple machine learning models to make sense of data that is dynamically webscraped. Each AI would handle a specific task, for example:

An AI model would identify text in an image, then attempt to create plain text of what it might be. Once the text is extracted, it would be passed in a stored variable to an AI that can read the text to determine if it is a US city/state.

I tried to look into if others have done this, but didn't find much on it relating to what I was looking for. Does anyone know if there are potential issues with this? Logically, it looks good to me, but I figured I'd ask.

If anyone can put me in the right direction for reading material or further information, I would appreciate it.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

You have explained a pipeline of AI algorithms for text in images: 1) Text detection, 2) OCR, 3) named entity recognition (NER). There are reams of paper on these topics.

Extracting City and Country Name from Text

Papers on Text from Images

Websites on Text from Images

Improve this answer
$\endgroup$
1
  • $\begingroup$ I think the keyword for OP is "pipeline". It is common practice to combine detectors, classifiers and other processing to create a pipeline. $\endgroup$
    – Neil Slater
    Commented Jun 3, 2021 at 12:55

You must log in to answer this question.