an error about bdd100k

[haozy008]

Thanks very much!
My issue is solved

[xinw1012]

Hi,

thanks for your interest in our work. Could you provide more details about the command you were using as well as the annotation labels?

When you download the latest BDD100K detection set from the website, you will see the labels in the BDD100K format named det_v2_{train,val,test}_release.json. To convert it, you can run python3 bdd100k/label/bdd100k2coco.py -i /path/det_v2_val_releast.json -o /path/det_v2_val_coco.json. For the training set, if you want to remove the additional classes (e.g., other person, other vehicle) in the annotations, you can add the flag --remove-ignore in the command.
I tested the script on my side and didn't find any errors.

Let us know if there is still an issue for you.

[sainisanjay]

@xinw1012 I was able to convert into bdd to coco. Could you help me, how we can convert into coco to Pascal VOC?