how to use mixup in detection model with keras?

[Jackwanguestc]

how to change the source code to fix detection, about two image size and box label add.
can it support keras Sequence, make the data generator to multi-processes version?

[yu4u]

I think it is easier to write dedicated codes from scratch for these purposes.
The data generator highly depends on object detection frameworks.