For my class, I've put together as many resources I could find (at least, in a few days of research), including every direct-downloadable dataset from the various crowdsourcing efforts and official places and investigative projects:

http://2015.padjo.org/briefs/tracking-police-involved-homici...

Counting police-officer-involved homicides is difficult -- but not more difficult than all the other things the FBI tries to count. That said, even if there were an official count, it's still very important that news orgs and independent groups also do their own count. Classifying these incidents gets significantly more difficult when you get into the details...but that's a problem common to every time we try to classify complicated information into a datapoint.