R.O.A.D. Barbados Historic Handwriting Challenge
'/%3e%3c/svg%3e)
Helping Barbados
$25 000 USD
Closing soon! (3 days left)
Optical Character Recognition
Natural Language Processing
1984 joined
749 active
Start
03 Jul 26
Close
04 Oct 26
Reveal
05 Oct 26
Hello, Zind,
5 Aug 2026, 11:40 · 7
I have CER 75% but reeaching 98+ by external datasets of relevant same hadwritten English. It is open public datasets. I see it is forbidden so will submit pure 4096 images by default. Is it possible to use external datasets which are opensource. Just a question ? I read rules but need to clarify this since 20000 images of relevant data are much better then 4000 . Now i am focused on 4000 to follow rules of competition but still interresting if relevant historical datsets are allowed/
@meganomaly Could you please calrify it I have same question here as well that if dataset is opensource and commericial usage is allowed Can we Use for Training our model
which external dataset is this?
i refused to use it as it is not clear if it is eligible or not. just used same century english from transcribus, Big enough structured same and stylistically identical. since i am not using it now results are not used for competiotion (TRCOR model jumped to the very high end of its capabilities) maybe qwen and ensemble with GPRO will do same as Jhonny mentioned.
yes just serach opensorce Something like Berigham... i dont remeber exact name.
Since it is not about competition i just did not used it. as far as i know even test dataset is not eligeible by rules yo train model so 4096 is what we are limited with withi9n BABADOS - check rules again.
thanks. was just curious.
NO , rules and req are very clear abt that.