Real-time Configuration Adaptation for Video Analytics

We explored the configuration adaptation with video analytics based on object moving features. We proposed a machine learning-based classification method to dynamic predict the configuration for future frames. We experimented on three applications: Pose estimation, Speaker detection and Traffic tracking with high performance.