colmap 을 사용 할 수 있는 hloc 도커에서 사용하기

 

1) 깃 레포 받기 

git clone --recursive https://github.com/cvg/Hierarchical-Localization/
cd Hierarchical-Localization/

2) 도커 빌드 

docker build . -t hloc:latest

(중 실패 )

 

이슈에서 다음을 찾음 https://github.com/cvg/Hierarchical-Localization/pull/330/commits/f3e43725cba31e75c23c95fdf08722d92c3c2347

Update Dockerfile for recent colmap by HelgeS · Pull Request #330 · cvg/Hierarchical-Localization

 

docker file 변경후 따라해봄 (+ apt-get install -y git)

FROM colmap/colmap:latest

MAINTAINER Paul-Edouard Sarlin

# ARG PYTHON_VERSION=3.8

# RUN apt-get update -y

# RUN apt-get install -y unzip wget software-properties-common

# RUN add-apt-repository ppa:deadsnakes/ppa && \

#     apt-get -y update && \

#     apt-get install -y python${PYTHON_VERSION}

# RUN wget https://bootstrap.pypa.io/get-pip.py && python${PYTHON_VERSION} get-pip.py

# RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTHON_VERSION} 1

RUN apt-get update -y && apt-get install -y unzip wget python3-pip && apt-get install -y git

COPY . /app

WORKDIR app/

RUN pip install --upgrade pip

RUN pip install -r requirements.txt

RUN pip install notebook

 

이제야 성공

 

3) 도커 런

docker run -itd --restart always --name hloc -p 8888:8888 -v /mnt/d/data:/usr/src/data hloc:latest /bin/bash

 

4) sfm.py 돌리는데,

hloc > extract_features.py 260 line 쯤에서 다음을 고칩니다. 

 

4-1) num_workers =0, pin_memory=False

 

4-2) sfm.py 에서 if __name__ == "__main__": 코드를 감싼다.

 

from pathlib import Path

from hloc import (

    extract_features,

    match_features,

    reconstruction,

    visualization,

    pairs_from_retrieval,

)

if __name__ == "__main__":

    images = Path("datasets/South-Building/images/")

    outputs = Path("outputs/sfm/")

    sfm_pairs = outputs / "pairs-netvlad.txt"

    sfm_dir = outputs / "sfm_superpoint+superglue"

    retrieval_conf = extract_features.confs["netvlad"]

    feature_conf = extract_features.confs["superpoint_aachen"]

    matcher_conf = match_features.confs["superglue"]

    # if not images.exists():

    #     !wget https://cvg-data.inf.ethz.ch/local-feature-evaluation-schoenberger2017/South-Building.zip -P datasets/

    #     !unzip -q datasets/South-Building.zip -d datasets/

    retrieval_path = extract_features.main(retrieval_conf, images, outputs)

    pairs_from_retrieval.main(retrieval_path, sfm_pairs, num_matched=5)

    feature_path = extract_features.main(feature_conf, images, outputs)

    match_path = match_features.main(

        matcher_conf, sfm_pairs, feature_conf["output"], outputs

    )

    model = reconstruction.main(sfm_dir, images, sfm_pairs, feature_path, match_path)

    visualization.visualize_sfm_2d(model, images, color_by="visibility", n=5)