How to achieve the federal image object detection task?

[MasterWinston]

First of all, thank you to the FATE contributors!
I have already implemented the mnist dataset classification task using the homo nn component (customizing CNN through the ModelLoader method), but I don't know how to handle the object detection dataset, upload it and train it. Do you have any relevant examples

Thanks again