monodepth2

Why we included this project

Depth from a single image is a hard problem, and this repo is the official PyTorch implementation of the self-supervised method described in the ICCV 2019 paper "Digging into Self-Supervised Monocular Depth Prediction." The key appeal is that it learns depth without ground-truth depth labels, so training does not depend on expensive depth sensors or manual annotation. For practical use, a single-image prediction script pulls down a pretrained model and returns disparity, or metric depth when you pick the stereo-trained weights, giving you a working estimate in minutes. The code is also the canonical release for researchers reproducing results on KITTI or adapting the pipeline to a custom dataset via the dataloader hooks. One thing to check before adopting it: the license is non-commercial, so it suits research and teaching rather than shipping products.