microsoft / MixedRealityCompanionKit

This is a MixedRealityToolkit style repository for code bits and components that may not run directly on Microsoft HoloLens or immersive headsets but instead pair with them to build experiences.
MIT License
591 stars 286 forks source link

SpectatorView: spatial mapping in compositor is different with the SpectatorPoseProvider #315

Closed cathtp6 closed 5 years ago

cathtp6 commented 5 years ago

I followed all the steps and SpectatorView is working generally. However, my hologram is not shown at the correct position in the compositor. The position of the hologram is only shifted a bit at the y-axis. We compare the live preview of the SpectatorView HoloLens and the spatial mapping in the compositor, the whole spatial mapping in compositor is higher compared with the reality. We redo the calibration and achieve the low RMS value, the translation matrix is also similar to the position where we mount the camera. Is the problem related to the calibration? We found that 2 months ago, there is a new calibration app in GitHub. The calibration data from this new app is much better. There are more parameters of the HoloLens in txt file but not been used in the script in the SpectatorViewSample. We also click the checkbox "Rotate Calibration" under calibration script in Unity. Does anyone know what else could I do or where I can check for? Appreciate for any help.

The setup we have: GoPro Hero5 using Linear mode, placed above the HoloLens. GoPro 1080P FPS:30 Unity 2017.4.1

chrisfromwork commented 5 years ago

Hello,

This is probably an issue with the physical transform you obtained through calibration. The accuracy of this process varies/it may take multiple attempts to obtain a usable transform. We are working on assessing what improvements can be made here. That going said, can you share a copy of your CalibrationData.txt? The translation and rotation values in this file may be able to indicate how accurate your calibration was if you physically measure distances from the HoloLens PV camera to the dslr camera. A hacky work around may be to update these translation values to account for the less than ideal y offset.

chrisfromwork commented 5 years ago

I am going to close this issue for now based on a lack of activity.

If you can share the contents of your calibration file, we may be able to better understand what is going wrong. That going said, this sounds like a calibration issue, and spectator view moving forward will have a different calibration solution.