search

Media-Smart / vedastr

A scene text recognition toolbox based on PyTorch
Apache License 2.0
533 stars 100 forks source link

Fine-tuning with custom dataset #73

Open chophilip21 opened 3 years ago

chophilip21 commented 3 years ago

Hello, I'm trying to test Vedastr for reading ID tags on animals (mostly cows).

Would it be possible to train on custom dataset? If so, what are the procedures?

I would like to use your pretrained weights to fine-tune on my custom dataset, and I am not sure where to look into exactly.

hungduong-vn commented 2 years ago

Yeah, I am in need of a scene text recognition module but fit on a Vietnamese dataset. Therefore, being able to fine tune the vedastr's models to different datasets would be nice since they are already trained on diverse and standardized datasets for scene text recognition task.

@ChaseMonsterAway Can you please provide us the steps to perform fine-tuning of your pretrained models?