RepeatFactorTrainingSampler for classes' bounding box instead of sample

[kwcckw]

Hi,

From the doc, it mentions a sample may appear more times than others so it is referring to the sample image?

https://detectron2.readthedocs.io/en/latest/modules/data.html#detectron2.data.samplers.RepeatFactorTrainingSampler

If i'm having batch size of 1, but each image has many bounding boxes with imbalance classes, will RepeatFactorTrainingSampler work the same in this case?

Thanks.