LKLM: A Large-Kernel Lightweight CNN Model for Monocular Depth Estimation

Depth estimation is a fundamental task in computer vision, which is increasingly widely applied in autonomous driving. However, many existing methods have often resulted in large computational overhead and high latency. To tackle this issue, we propose a lightweight CNN architecture named Large-Kernel Lightweight CNN Model (LKLM). Extensive experiments demonstrate the superiority of our approach in lightweight depth estimation. On the KITTI dataset, LKLM achieves a large margin in accuracy, with 4.254 GFlops.