Questions regarding generation and recognition task

[staghado]

Hi there,

thank you for your work!

I have trained the model with a similar setup on IAM dataset. I have a few questions for you: 1- is it normal that for some styles the model makes errors like forgetting some characters in the generated image? Is it due to poor coverage of the training dataset ; for example for some styles the model has never seen some letters during training. 2- how much time would it take to generate the whole IAM dataset (around 45k images) with random styles on a gpu(a100 or v100) ? I m asking this because I think you have already done this. 3- for the OCR task : how much data was generated to perform the recognition task?

thank you in advance for your reply! Said

[staghado]

Any updates on this?

[Fyzjym]

The same question.