Benchmark for anonymous video analytics

Abstract Out-of-home audience measurement aims to count and characterize the people exposed advertising content in physical world. While solutions based on computer vision are of increasing interest, no commonly accepted benchmark exists evaluate compare their performance. In this paper, we propose first for digital out-of-home that evaluates vision-based tasks localization counting, demographics. The is composed a novel, dataset captured at multiple locations set performance measures. Using benchmark, present an in-depth comparison eight open-source algorithms four hardware platforms with GPU CPU-optimized inferences two commercial off-the-shelf localization, count, age, gender estimation. This related codes available http://ava.eecs.qmul.ac.uk .