Skip to content

Document that exported poses are metric scale (closes #8) #18

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
kunjikakgp wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Document that exported poses are metric scale (closes #8) #18

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,6 +97,8 @@ The `t_ij` element of the transform matrix is in row-major order, so `i` is the

**NOTE:** The coordinate system of the camera poses follows the ARKit gravity aligned convention which you can read about [here](https://developer.apple.com/documentation/arkit/arconfiguration/worldalignment/gravity). In short, this is a right-handed coordinate system where the `+Y axis` points up, and `-Z axis` points in the direction that the camera was pointing when the session started, and perpendicular to gravity. Note: depending on the coordinate system convention of your downstream application you might have to apply an additional rotation to the camera transform to get the conventions to match.

**Are the poses metric scale?** Yes. ARKit's camera tracking fuses the device's IMU (accelerometer) with visual data via visual-inertial odometry, and on LiDAR-equipped devices further grounds tracking with direct depth measurement -- both give the exported poses a real-world physical unit reference, so translations are in meters. This is different from vision-only SfM pipelines like COLMAP, where scale is ambiguous unless recovered from an external reference (a measured object, GPS, or fused IMU data) -- one of the practical advantages of using Polycam's raw export for applications like metric-scale NeRF/3DGS training. (See issue #8.)

### corrected_cameras

This directory includes the improved camera poses that have been globally optimized. The data is formatted in the same way as the cameras described above, and you should almost always use these files intead of the original camera files. If this directory does not exist, it means the session was too large for the pose optimization step to run on device, but you can still fall use the unoptimized ARKit camera poses.
Expand Down