In this work, we propose the first end-to-end optimized framework for compressing automotive stereo videos (i.e., stereo videos from autonomous driving applications) from both left and right views.