This disclosure provides a method performed by a portable computer device configured to generate an augmented reality video. The device detects first object proposal region information, based on a frame of a video, using a first trained model configured to provide object proposal regions having an accurate width. The device also detects, based on the frame of the video, second object proposal region information using a second trained model configured to provide object proposal regions having an accurate height. The device then determines combined object proposal region information by combining overlapping object proposal regions of the first and second object proposal region information and generates an augmented reality video by generating an augmented frame. The augmented frame is generated by overlaying object proposal regions that are in the combined object proposal region information onto the frame of the video and adding the augmented frame to the augmented reality video