Hi,
I am currently using the Gemini Flash API for OCR tasks, but I am experiencing accuracy issues. While AI Studio provides over 99% OCR accuracy, the same prompt and a 60KB image only achieve around 70% accuracy via the API. I suspect that the image quality might be significantly reduced internally within Gemini. Does anyone have insights or suggestions on this matter?
Thank you.
same problem here, around 75% accuracy with the API and 98% with AI studio
User | Count |
---|---|
2 | |
1 | |
1 | |
1 | |
1 |