SuperPoint replace the sift in colmap framework
- install the colmap: https://colmap.github.io/install.html
- clone the super colmap: git clone https://github.com/Xbbei/super-colmap.git
- Download SuperPoint to ./
wget https://github.com/magicleap/SuperGluePretrainedNetwork/blob/master/models/weights/superpoint_v1.pth?raw=true -O superpoint_v1.pth
- Python 3.7+
- PyTorch and TorchVision
- OpenCV
- imageio and imageio-ffmpeg
python super_colmap.py \
--projpath projpath \
--cameraModel SIMPLE_RADIAL \
--images_path rgb \
--single_camera
projpath is your projpath, it must have the "rgb" or "images" dir that contains the images; cameraModel can be "SIMPLE_RADIAL, SIMPLE_PINHOLE, PINHOLE, RADIAL, OPENCV, FULL_OPENCV, SIMPLE_RADIAL_FISHEYE, RADIAL_FISHEYE, OPENCV_FISHEYE, FOV, THIN_PRISM_FISHEYE"; images_path is rgb and it denotes that the images is in projpath/rgb; single_camera controls the images are in the same model.