How to use ESOD framework with YOLOv8

Hi,
Thanks for your great work on ESOD: Efficient Small Object Detection on High-Resolution Images.

I’d like to train YOLOv8 models within the ESOD framework.
From the README, I noticed that YOLOv8 does not require model conversion, but there are no detailed instructions yet.

Could you please provide some guidance or examples on:

How to correctly integrate YOLOv8 weights and configs into ESOD?

Whether any modifications are required in train.sh, model_convert.py, or the data configuration files?

If there are compatibility issues between YOLOv8 and the ESOD training pipeline (e.g., image tiling or patch merging)?

Any short example or clarification would be greatly appreciated.
Thanks again for sharing this excellent work!

Start with the README and inspect train.sh, model_convert.py, and the data configuration files to understand the existing model workflow. Document how YOLOv8 weights and configs fit into ESOD, including any required changes and compatibility notes for image tiling or patch merging. Done means a short, reproducible YOLOv8 integration example is available.

Written by the indexing model from the issue text.