I'm currently working on a project where I need completely accurate user data, with no room for error. I'm looking for a tool or service that can analyze photos of IDs, bank cards, and other official documents. These images may vary in quality, angle, rotation, and could have some minor scratches. Since my entire production stack is already in Azure, I'm wondering if Azure Document Intelligence is the best solution for this task. What are the expected response times for extracting information from three such images? I'd love any advice or suggestions you might have!
5 Answers
Security should definitely be a top priority for you. There are huge risks if sensitive data falls into the wrong hands.
There's no such thing as perfect OCR. Azure Document Intelligence is solid, especially since you're already using Azure for everything else. Just keep in mind that it won't be flawless, so don't expect 100% accuracy.
You should know that achieving 100% accuracy is not realistic; it's something that just doesn’t exist in tech. Even human-based recognition has its flaws. Don't put too much pressure on needing it to be perfect.
I totally agree with the idea of aiming for 100%. You should understand that nothing can guarantee it, but AI and OCR are typically more reliable than human accuracy.
Just a heads up, you'll need to train the model if the documents you’re using have a lot of variation. It’s something to think about when setting this up.
True! But, since my app will be for contract creation, I need the data to be legally accurate. I'm trying to minimize the error rate as much as possible.