Not sure how relevant those results are for CCTV. Looking at the pictures in the dataset (http://megaface.cs.washington.edu/dataset/subset.html#facesb...), these are nowhere close to what CCTV looks like.