I work in optics and computer vision, and investigate research papers on a regular basis. In the majority of the papers I read the methods proposed either only work for the specific example in the paper (some don't even do that), or only work under such "ideal" circumstances that they barely work under lab conditions, much less real world cases. There are a few gems, but you can tell that much of what is published is just written so someone gets that degree or that quota to reach tenure. Prior to working at this job, my trust in published research was much higher. Now my tendency is to treat it all as garbage until it proves itself otherwise.