I was trying to read ugly handwriting, and it seems to me that Gemini 1.5 OCR is producing better output than Gemini 2.0 Flash. I just want to know if 1.5 uses better OCR technology or what the reason could be for the difference in output quality.
Hi @Syeed_Talha, If possible could you please provide the sample image that you are trying to extract the text from the image where you have observed the performance difference between 1.5 and 2.0 flash model. Thank You.
Here is a sample image:
Here is the system prompt you can try:
Read the medicine name and dosages very carefully and give me the perfect output .
Output format:
Medicine 1:
Medicine name:
Dosage: 1+0+0
Medicine 2:
Medicine name: Napa
Dosage: 1+1+0
you can try this image to extract medicine name and dosage. It seems to me 1.5 pro perform well.
Hi @Syeed_Talha, The Preprocessing for the input data remains same for all the model. Thank You.
1 Like
I actually did some ocr (before 2.5) and I personally think that 1.5 Pro is better than 2.0 Flash.
1 Like