This is the official implementation for training and testing depth estimation using the model proposed in

HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation

Xiaoyang Lyu, Liang Liu, Mengmeng Wang, Xin Kong, Lina Liu, Yong Liu*, Xinxin Chen and Yi Yuan.

This paper has been accepted by AAAI 2021.

Note: We temporarily release the evaluation version and some pretrained models of our paper. The training codes are modified according to Monodepth2, and we will release them soon.

Update

2021.1.27

1. The training code will be released around the beginning of the March.
2. For re-implementing HR-Depth, you can clone Monodepth2 and simply replace the DepthDecoder with HRDepthDecoder. Our parameter settings are exactly the same as Monodepth2.
3. In our paper, we wrote the initial learning rate wrong. It should be 1e-4, not 1e-3. We will fix this mistake in the final version. Thanks for someone pointing out our problem.

Quantitative Results

HR-Depth Results

Lite-HR-Depth Results

Usage

Requirements

Assuming a fresh Anaconda distribution, you can install the dependencies with:

conda install pytorch=1.5.0 torchvision=0.6.0 -c pytorch
conda install opencv=4.2
pip install scipy=1.4.1

Pretrained Model

We provided pretrained model as follow: