I am working on fine-tuning a detector using a custom dataset in Sanskrit (handwritten manuscripts). Since the ICDAR 2013 dataset utilizes character-level annotations, I wanted to confirm if I also need to provide character-level annotations for other languages in my dataset. Since I previously provided word level annotations in finetuning EasyOCR which also uses CRAFT model only for Text Detection
Hi @faustomorales,
I am working on fine-tuning a detector using a custom dataset in Sanskrit (handwritten manuscripts). Since the ICDAR 2013 dataset utilizes character-level annotations, I wanted to confirm if I also need to provide character-level annotations for other languages in my dataset. Since I previously provided word level annotations in finetuning EasyOCR which also uses CRAFT model only for Text Detection
Thank you!