People are excellent at recognizing each other's emotions in general, but with the development of artificial intelligence and robots, it's critical to investigate if computers can perform emotion recognition as well as humans in real-time. The goal of the research is to make the performance of emotion recognition as good as possible in real-time on a local processor. To optimize computational speed, the real-time pipeline from taking images to output predicted emotion was created on the local MacBook and low-cost Raspberry Pi 4. Results from the laptop prove the speed increased considerably compared to the previous cloud platform. Furthermore, a human subject experiment was designed to figure out whether the accuracy of the model would still perform a good result with different distance settings.