modyfing YOLOv8 architecture

[cssaroon]

Hi

I’m currently working on a research project and would like to customize the YOLOv8 architecture—for example, by modifying the backbone and neck by adding attention mechanisms like (CBAM) to enhance the model performance and solve the problem of occlusion in the process of detecting heads in a crowded scene. Could you please guide me on how to properly edit the architecture in the source code (e.g., which files and classes to look into)?

Any documentation, tips, or references would be greatly appreciated!

Thanks in advance!

Best regards,