A system may be configured to perform object counting in high volume traffic. In some aspects, the system may detect a candidate object within the region of interest in a current video frame, and determine that the candidate object is a detected object based at least in part on comparing an attribute value of the candidate object to historic attribute information determined during a plurality of previous video frames. Further, the system may determine track information based on the detected object and determine an object count representing a number of the objects that have entered the region of interest and/or a number of the objects that have exited the region of interest.