The testing results of the whole dataset is empty about swin-transformer-object-detection HOT 5 OPEN

I also have this error for the first 10 epoch, but it stopped after 10 epoch.Is your MAP value normal?

from swin-transformer-object-detection.

if you are training object detection without mask, your annotation area might be 0, the below check will cause no data input.

change the mmdet/datasets/coco.py line146
if ann['area'] <= 0 or w < 1 or h < 1
to if w < 1 or h < 1

from swin-transformer-object-detection.

I'm very glad to receive your reply. Thank you for proposing such an excellent swin transformer。
but
After the modification, there is still this error when running the test command. Is there anything else to modify. My config_file is cascade_mask_rcnn_swin_tiny_patch4_window7_mstrain_480-800_giou_4conv1f_adamw_3x_coco.py

from swin-transformer-object-detection.

it look very similar to this issue
open-mmlab#2942 (comment)

from swin-transformer-object-detection.

Thank you, your solution works. The problem is related to the lack of mask.

from swin-transformer-object-detection.