Existing SLAM (Simultaneous Localization and Mapping) algorithms have achieved remarkable localization accuracy in dynamic environments by using deep learning techniques to identify dynamic objects. However, they usually require GPUs to operate in real-time. Therefore, this paper proposes an open-source real-time dynamic SLAM system that runs solely on CPU by incorporating a mask prediction mechanism, which allows the deep learning method and the camera tracking to run entirely in parallel at different frequencies. Our SLAM system further introduces a dual-stage optical flow tracking approach and employs a hybrid usage of optical flow and ORB features, enhancing efficiency and robustness by selectively allocating computational resources to input frames. Compared with previous methods, our system maintains high localization accuracy in dynamic environments while achieving a tracking frame rate of 56 FPS on a laptop CPU, proving that deep learning methods are feasible for dynamic SLAM without GPU support. To the best of our knowledge, this is the first SLAM system to achieve this.