Image IDs missing in Mixed dataset

[aitor-martinez-seras]

First of all, thank you for your outstanding work and for sharing it!

I have encountered a problem with the image IDs of the annotations of the mixed dataset. To make it clear in my explanation, this is the structure of the annotations I'm referencing:

loaded_json = {
 'annotations': [{... 'image_id': int ...}]
 'images': [{... 'id': int,...}]
 ...
}

The problem is the following: the annotations of instances_val2017_mixed_OOD.json (inside the annotations dict, in the image_id key) refer to image ids not present in the images dict (in the id key). This image ids referenced in the annotations dict are from the validation split, but as mentioned are not included correctly inside the images key.

I'm assuming that this is handled in the lines 87 to 93 in the file aose.py, therefore the ids of the images not contained in images I assume that are not going to be included for computing the metrics (they are 62 ids I think). Confirm me please if this is the case.