Whilst there are many problems with "eye witness" accounts at least the witness can be cross-examined in court.
Furthermore, different people can provide independent corroborating or challenging evidence (provided they have not conferred beforehand) - though there could be common cognitive and perception issues that affect results (the classic 'Did you see the gorilla' experiment is an illustration of this).
It seems to me that judges/juries are often not well informed on the reliability of either eye witness or computer based evidence (or several other types of evidence).