3QFP

3QFP: Efficient neural implicit surface reconstruction using Tri-Quadtrees and Fourier feature Positional encoding [ICRA24]

Arxiv

Overview

Overview of our method.

Installation

The code is based on the implementation of nice work SHINE-Mapping, so we suggests the same way to install:

• Create a conda environment:

conda create --name shine python=3.7
conda activate 3qfp

• Install torch-related things

pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu116
pip install kaolin==0.12.0 -f https://nvidia-kaolin.s3.us-east-2.amazonaws.com/torch-1.12.1_cu116.html

Other torch versions should also work.

• Install other dependencies

pip install open3d scikit-image wandb tqdm natsort pyquaternion pyyaml torch_scatter==2.1.0
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch # for ff encoding

Dataset

Also, similarly, we suggest the download scripts from SHINE-Mapping.

• MaiCity dataset

sh ./scripts/download_maicity.sh

• KITTI dataset

sh ./scripts/download_kitti_example.sh

• Newer College

sh ./scripts/download_ncd_example.sh

In the configuration (.yaml) files, you can specify the dataset path.

pc_path: the folder containing the point cloud (.bin, .ply or .pcd format) for each frame. pose_path : the pose file (.txt) containing the transformation matrix of each frame. calib_path : the calib file (.txt) containing the static transformation between sensor and body frames (optional, would be identity matrix if set as '').

Run

python run.py ./config/maicity/maicity_batch.yaml