Computation of TPs and FPs in COCOeval class

[CRISBAKH]

In the COCOeval class documentation, we find:

dtMatches - [TxD] matching gt id at each IoU or 0

gtMatches - [TxG] matching dt id at each IoU or 0

What if a matching happens for the first dt of gt objects that have 0 index ? Then a 0 in dtMatches or gtMatches will be wrongly interpreted during TP and FP computation no ? I'm talking about case like this

dtMatches = [1.0, 2.0, 0.0, 0.0, 0.0, 0.0, 0.0] --> we have 7 available detections with indices dtIds = [0, 1, 2, 3, 4, 5, 6]
gtMatches = [3.0, 0.0, 1.0, 0.0] --> we have 4 available ground truth with indices gtIds = [0, 1, 2, 3]

if we check both dtMatches and gtMatches, it seems that dt_matches[3]=0 is not a non matching indicator but more a matching with the gt object of index 0 where gtMatches[0] = 3. Then how to differentiate a 0-index matching from a 0 reflecting a non matching found ?