How do I fine-tune a model using my own datasets in Automatic Labeling( RAM or Tag2Text )

[leershavin]

The pth file of DINO has been trained and fine-tuned, and it can be called normally under the keyword prompt mode.

Now I want to use Tag2Text or RAM’s full automatic annotation. Do I need to retrain and fine-tune? If so, can you provide the method for training and fine-tuning?

[ShuoShenDe]

Hi dear, If you want to use Tag2Text or RAM’s full automatic annotation and it meets your data requirements, there's no need to retrain or fine-tune the model. However, if the current solutions do not fully meet your needs, you may consider fine-tuning.

While official fine-tuning guidelines are not provided, some teams have developed their own solutions. You can explore these approaches by referring to the relevant repositories, such as this one: MMDetection Grounding DINO.

Susan Shen