This repository provides the official implementation of "STATIC: Surface Temporal Affine for Time Consistency in Video Monocular Depth Estimation". STATIC enhances video depth estimation by independently modeling static and dynamic regions through surface-normal-based masking, achieving state-of-the-art temporal consistency without relying on optical flow or camera parameters.
You can download the pretrained models "kitti" from here ckpt