The repository is the code implementation of the paper A Two-Stage Masked Autoencoder Based Network for Indoor Depth Completion, based on MAE projects.
If you find this project helpful, please give us a star ⭐️, your support is our greatest motivation.
- Introduction
- Installation
- Dataset Preparation
- Model Training
- Model Testing
- Image Prediction
- Acknowledgements
- Citation
- License
- Contact Us
- Ubuntu
- Python 3.7+, recommended 3.7.0
- PyTorch 1.9.0 or higher, recommended 1.9.1+cu111
- CUDA 12.4 or higher, recommended 12.4
We recommend using Miniconda for installation. The following command will create a virtual environment named idc
and install PyTorch.
Note: If you have experience with PyTorch and have already installed it, you can skip to the next section. Otherwise, you can follow these steps to prepare.
Step 0: Install Miniconda.
Step 1: Create a virtual environment named ttp
and activate it.
conda create -n ttp python=3.7 -y
conda activate idc
Step 2: Install PyTorch2.1.x.
Linux:
pip install torch==1.9.1 torchvision==0.10.1 torchaudio==0.9.1 --index-url https://download.pytorch.org/whl/cu111
Step 3: Install [timm]
pip install timm=0.4.9
Step 4: Install other dependencies.
pip install matplotlib scipy numpy opencv-python pillow typing-extensions=4.2.0
Download or clone the repository.
git clone [email protected]:kailaisun/Indoor-Depth-Completion.git
cd Indoor-Depth-Completion
Image and label download address: Matterport3D for Depth Completion. It includes:
- train_full : A training dataset of npy files which is concatenated from rgb images, raw depth images and gt depth images for finetuning.
- test_full : A testing dataset of npy files which is concatenated from rgb images, raw depth images and gt depth images for finetuning.
python main_pretrain.py --data_path /npy/train_full
python main_fintune_full.py --data_path /npy/train_full --eval_data_path /npy/test_full
-Pretraining:Download
-Finetuning:Download
python eval_full.py --data_path /npy/test_full --checkpoint /checkpoint-finetune.pth --output_dir /output # data_path is the file to be tested, checkpoint is the checkpoint file you want to use, output_dir is the output path of the prediction result, including predicted depth images and point clouds.
The repository is the code implementation of the paper A Two-Stage Masked Autoencoder Based Network for Indoor Depth Completion, based on MAE projects.
If you use the code or performance benchmarks of this project in your research, please refer to the following bibtex to cite.
@misc{sun2024twostage,
title={A Two-Stage Masked Autoencoder Based Network for Indoor Depth Completion},
author={Kailai Sun and Zhou Yang and Qianchuan Zhao},
year={2024},
eprint={2406.09792},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
The repository is licensed under the Apache 2.0 license.
If you have other questions❓, please contact us in time 👬