What is the average response time for your APIs?
Response time depends on the following parameters :
- is it an image (faster) or a pdf file (slower) ?
- What is the image resolution ?
- Number of data points returned by the API you are calling
Overall, on average, response time is below one second for a medium-resolution image (a few Mpx) and can goes up to a few seconds a page for a large pdf file.
What is the average accuracy of extraction?
Before releasing any API to production, we compute extraction accuracy and precision on a large dataset (~10k data) as diverse as possible.
None of our algorithm goes into production before we get at least 90% of accuracy for each field individually.
Depending on the field complexity, you should expect accuracy to be between 90% and 100%.
Note that if you compute accuracy below 85% for a specific field on a production-like dataset, this is highly possible that something is wrong with your test pipeline.
To make sure your tests went the right way and your metrics can be trusted, feel free to ask the chat to talk to an expert, we are all the more interested in your feedback to understand what went wrong in order to make our algorithms better.
Is there any technical limitation?
In order to protect our infrastructure, the following technical limitations are implemented on all our APIs:
- A maximum number of calls to be made per second, per minute and per hour. By default, the limit is set at 4 per second, 100 per minute and 3000 per hour
- A maximum resolution for images set by default at 15MPixels
- A maximum number of pages for a pdf file set by default at 3 pages
What can I do if I need to extract information from a document that is not yet supported?
If you need to extract information thanks to a dedicated API, please tell us about your use case and the document you need to parse to see whether we could build it just for you.
If you need to build a smooth visual interface to extract and/or validate the information extracted from a document, you can use our open source frontend SDK and Mindee Vision API (github here) to build your dream UI for any document.
Can I invite other people from within my organization to the Mindee platform?
You can add as many users as you want into your workspace. To do so, find the member section in your account settings on the platform.
I did not receive my confirmation email, what should I do?
Sometimes, confirmation emails land into the spam section of your mailbox. Go check out yours to make sure it did not happen.
If you still cannot find it, ask the chat to get a new one. Someone from our team will send one right away.
Can I send multiple page pdf?
Yes. We support mutiple-page pdf files. Our APIs handle each page as a single image and return a prediction for each page. If you send a three-page document, you will get three predictions.
Can I send a pdf containing several documents?
Yes you can but we will not be able to cut it for you into several documents. We do not provide any document processing feature, only document understanding.
If you send a pdf file with multiple pages, we will make a prediction for each page.
Do you support handwritting?
We do not support handwritten documents for the moment, but it's scheduled in our roadmap.