some problem with the builtin file

[captain0305]

Hi! This work is quite amazing. And I have some problems with the file builtin. I don't understand why there's a domain_path. And the file "bdd100k_mot_domain_splits_train.json" is different from the file convert "bdd100k/labels/track/bdd100k_mot_train_coco.json" or not?

How could I generate the file under the domain_path?

`

_PREDEFINED_SPLITS_BDDT = {
    "bdd_tracking_2k": {
        "bdd_tracking_2k_train": (
            "bdd100k/images/track/train",
            "bdd100k/labels/track/bdd100k_mot_train_coco.json",
        ),
        "bdd_tracking_2k_val": (
            "bdd100k/images/track/val",
            "bdd100k/labels/track/bdd100k_mot_val_coco.json",
        ),
    },
}

# Register data for different domains as well as different sequence.
domain_path = "bdd100k/labels/box_track_20/domain_splits/"

train_splits = load_json(
    osp.join("/nobackup-slow/dataset/my_xfdu/video/bdd/", domain_path, "bdd100k_mot_domain_splits_train.json")
)
val_splits = load_json(
    osp.join("/nobackup-slow/dataset/my_xfdu/video/bdd/", domain_path, "bdd100k_mot_domain_splits_val.json")
)

`

[captain0305]

I've solved this issue by using this file

[yeyang1021]

I've solved this issue by using this file Can you provide the specific process ?

[jadevaibhav]

@yeyang1021 Can you please explain how you are able to resolve this issue? Also, @d12306 can you please provide the exact detectron2 library version you used? I am getting errors loading the SSRCN model (from stud_resnet_ood_coco.yaml). Thanks!