Open Knivacke opened 9 months ago
It's better to finetune Grounding-DINO for better localization in this pipeline
Thanks for the response. As in fine-tuning Grounding-DINO separately, and using my fine-tuned model in place of base grounding DINO for grounded SAM?
First of all - what an amazing framework, I'm blown away.
Is it possible to fine-tune the automatic labelling process of grounded SAM, ie. grounded DINO? I'm using the model on newspaper images from Sweden, and would for example need more consistent classification of Swedish police officers.
Thanks.