The Innocence Project has used DNA and other forensic evidence to exonerate people convicted primarily on the basis of eyewitness testimony. Juries tend to give eyewitness testimony for more weight than it deserves.
The Innocence Project has been successful in setting aside convictions based on mistaken eyewitness identification evidence. Eyewitness identification evidence is not fact observation unless the witness knows the person and recognizes them at the time of the offence. There are also serious problems with ensuring the evidence is not contaminated with bias of the official conducting a photo or in-person line up.
But the Innocence Project has also been successful in reversing convictions by getting forensic evidence thrown out: Good examples being faulty bite mark evidence posing as “science”; comparative bullet lead analysis being rejected completely; rejecting evidence of hair and fibre comparison “matches”; and getting opinions of so called arson experts excluded (have a look at the tragic Texas case in which Todd Willingham was convicted to see what kind of junk science masquerades as “forensic science”.).
In Canada, the wrongful conviction of a man charged with murder based on flawed forensic medical evidence that conflicted with the timeline established by several independent witnesses was set aside after a reenactment showed how the victim likely died. The correct verdict showed that the witnesses were right. Lookup the Clayton Johnson case.