About detection task

[QY1994-0919]

Hello, I am interested in your task, especially the application of droppos in detection and segmentation tasks, may I ask if you can share the code about detection task separation

[Haochen-Wang409]

Hi, thanks for your attention!!

For detection and segmentation, we simply follow the configuration of ViTDet produced in detectron2.
To train models initialized by DropPos, you need to modify the train.init_checkpoint in https://github.com/facebookresearch/detectron2/blob/main/projects/ViTDet/configs/COCO/mask_rcnn_vitdet_b_100ep.py#L19.