Annotations are for the observation itself, not any specific pictures, so it is possible that observations given the time and place when an organism was alive can include pictures taken later when the organism is dead. Such observations should be annotated as alive.
If the date/time of your observations are after death, then I would annotate dead.
For the observations you mentioned, assuming the time of the observation is when the fish were alive, I’d annotate as alive. On a related note, this is probably more informative about the species as they definitely were alive at the time observed (and most users of data will be looking for “alive” records). Users might often assume that “dead” records shouldn’t be used/mixed in with live ones.
There’s also some related discussion on this topic:
https://forum.inaturalist.org/t/dqa-for-specimens-removed-from-the-wild/40368