Question on Comparative Experiment Settings

[Kkubuck]

Hello,
I would like to ask about the settings for the comparative experiments.
In the Castdet comparison experiment, how were the base class dataset of DIOR and the NWPU VHR-10 dataset used when training YOLO-World?

Was YOLO-World trained from scratch using the author’s DIOR and NWPU datasets in addition to the known O+G+C datasets? Or how was the fine-tuning conducted?

Thank you for your response!