How can you integrate OpenCV with a deep learning model to perform real-time object detection, and what techniques would you use to optimize its performance?