the meaning of --keep_res?
i found that without --keep_res, the performance is better. what does --keep_res mean?should i add the option --keep_res
when i trained model on the dataset made by me, i added the option --keep_res, but when i tested it, i found that the ap is higher without --keep_res. so in my paper , which one should i use? please help me
would you please tell me why add the option -keep_res when testing while not add when training?

[keep_res](https://github.com/xingyizhou/CenterNet/blob/master/readme/GETTING_STARTED.md#coco) says
--keep_res is for keep the original image resolution. Without --keep_res it will resize the images to 512 x 512