search

colmap / colmap

COLMAP - Structure-from-Motion and Multi-View Stereo
https://colmap.github.io/
Other
7.82k stars 1.53k forks source link

About --ImageReader.single_camera_per_folder #1075

Open ghost opened 3 years ago

ghost commented 3 years ago
Hi, thank you always for your support. I consider the following situation: 1) There are two cameras handled by two persons. 2) They move around independently. 3) So, the input path has the following structure: [Input path] _camera1 _image001 _image002 _...
_camera2
_image001
_image002
_...

4) I need to specify intrinsic parameters for each camera, and optimize those two intrinsic parameters for the two cameras as well as poses simultaneously.

For that purpose, is the following command makes sense?

colmap feature_extractor \ --database_path [path]/database.db \ --image_path [input_path] \ --ImageReader.single_camera_per_folder 1 \

I looked your tutorial about "--ImageReader.single_camera_per_folder", but there seems difficult to understand how it specifies. I wonder, "--ImageReader.single_camera_per_folder 1" means "each camera folder has a single set of intrinsic parameters and can be optimized while reconstruction", doesn't it?

Thank you in advance.

TroiiF commented 3 years ago

Yes, you are right. The structure of images folder should be like this: -| images_folder --| camera1(sub folder1 of images_folder) ----| image1.jpg ----| image2.jpg ----| ... --| camera2(sub folder2 of images_folder) ----| image1.jpg ----| image2.jpg ----| ...

moshebou commented 1 year ago

Hi, A followup question - how can I specify the intrinsics parameters for each folder in the command line?

MaxPalmer-UH commented 1 year ago

In this case will it use the database with camera intrinsics specified there for each camera ID? If there is no database, how do you supply the intrinsics for each camera as @moshebou says? @TroiiF

MaxPalmer-UH commented 1 year ago

Also, how does it know which camera applies to each folder?