Due to the differentiability of the C2P conversion, our network autonomously learns planar visibility confidence by constraining the estimated plane parameters and reconstructed planar depth map. We ...