Video Detection (VD) systems for intersection signal operation are becoming morepopular and some agencies are using the counting feature of the VD system for trafficvolume count. In this paper the accuracy of the VD systems for traffic counting isevaluated. Counts obtained from two VD systems are compared with the actual trafficcounts in cloudy day and night conditions. Multiple datasets are used for comparison. Forboth the systems, it was found that the VD system's error is lower under cloudyconditions. Under night conditions the counts are consistently higher because of the falsecalls caused by the headlights of the vehicles in the adjacent lanes. It is recommended touse a longer time interval for the counts to reduce the range of error that exists for shorttime period counts.