JaidedAI
easy ocr does not work in docker deployment
Hi,
I am trying to deploy and use EasyOCR in a customized Docker container. When I run my PNG in the Docker container, the OCR only recognizes various empty strings like '', '', etc. Locally on my Mac M1, it works perfectly! I have already checked the library dependencies and am slowly getting desperate - does anyone have an idea which library could be causing this?
Hi, I have the same problem. I was developing the project on macOS/Windows and everything worked well — it recognized characters very accurately. But when I tried to package it into Docker, EasyOCR started recognizing only a few words from an image full of text (I had tested it on macOS where it was able to extract a lot of text, but after setting it up in Docker, it recognized only a few words).
Hi,
I am trying to deploy and use EasyOCR in a customized Docker container. When I run my PNG in the Docker container, the OCR only recognizes various empty strings like '', '', etc. Locally on my Mac M1, it works perfectly! I have already checked the library dependencies and am slowly getting desperate - does anyone have an idea which library could be causing this?
It turns out that if you run the same Docker configuration on Windows, EasyOCR works perfectly. But if you run it on macOS, for some reason it only extracts a single line. I suspect that my Mac has limited resources, which aren’t enough for the container to run properly.
I can also confirm the same problem on macOS with M1. Sometimes only a few rows are returned. And sometimes some trash like "L=LELELELELELELEEIEIE=I GLELELELELELE" (see attachments)
I double checked: -Python version and all dependencies -Base docker image (tried different) -Downloading language models in EasyOCR -Resources allocated for the Docker container -GPU & CPU-only modes
I'm having the same issue. Just like you with the LELELE, etc. Has anyone found a solution?
Update: setting quantize to False seems to do it for me
