About the comparison between swin-t and r50

[W-QY]

Dear authors,
Thanks for your great work！

I'm recently training some existing detection models with different backbones using my dataset in coco format. I tried to modify the r50 backbone of the ATSS model according to "configs/swin/mask_rcnn_swin-t-p4-w7_fpn_1x_coco.py", but the final result showed that there was no significant difference between the detection results of swin-t and r50. I would appreciate it if you could tell me the possible reason.