I am using the pro version of OCR API. I am running a script where I am using multithreaded environment and trying to get the text back from API with 14-15 threads. Most of the times, the requests get errored out. I am definitely not hitting the 100 images/second mark here. Can someone please help with this? For eg - I ran a dataframe having 1000 images and using an environment with 12 CPU. Almost half of the requests did not return text and got errored out.
Wow, that is a lot! Are you on the standard PRO plan? Then you are certainly hitting the rate limit of this plan.
Technically speaking, even without rate limit, the key limitation of cloud OCR is the image upload speed. So real time OCR on a video is not possible, no provider offers that.
Having said that, we offer local OCR. This avoids the image upload limitation, since it runs directly on your system.
so what exactly is the limitation on multiple simultaneous connections?
Did not find anything in the docs. Just the 500/day limit for free plan.
The free plan allows 2 simultaneous connections.
For the PRO OCR tiers, it depends on the plan. But it will be at least 5 connections to the fast PRO endpoints.