3DF Zephyr v.1.007

[Roberto]

Changelog:

- Fixed an issues during 3DF Samantha matching phase (improved accuracy)
- Improve speed of 3DF Samantha matching phase
- Fixed progress bar ETA
- Speed improvements during keypoint extraction
- Added Educational (for courses) version

[Roberto]

[!--:en--]The command line version 1.60 matches the one running in Zephyr v.1.007.

- Improved speed and memory footprint with large datasets
- Improved matching accuracy and speed
- Improved autocalibration[!--:--][!--:it--]The command line version 1.60 matches the one running in Zephyr v.1.007.

- Improved speed and memory footprint with large datasets
- Improved matching accuracy and speed
- Improved autocalibration[!--:--]

[Roberto]

I've uploaded a modified versions at this address:
x64
x86
The fix should work, but I haven't had time to test the new version properly yet.
Let me know if you are having problems.

[jamesm]

Thanks Roberto.
I'm getting the tangential components reported now, but error (as given by image residuals on control point observations) has increased significantly. This could be an issue here though - I will check further and compare with my v1.5 results.
All the best,
Mike

[jamesm]

Looks like the errors were a graphics card issue on the laptop - results look fine when processing was carried on a different machine.
Thanks,
Mike

[jamesm]

Hi Roberto,
I am trying to use the new Samantha v1.60 (x64), but the downloaded file (19,426,711 bytes) won't unzip (the archive appears to be corrupt). The x84 version seems fine. Any suggestions?
Many thanks,
Mike

[Roberto]

Hi Mike,
Thanks for reporting this. I've made an update yesterday evening to include a missing dll and it probably got corrupted during the upload. It should be ok now (size 19.618.399 bytes)

[jamesm]

Perfect. Thanks for the speedy response, it's running fine now.
Mike

[jamesm]

Hi Roberto,
Another query - which I think maybe a v1.60 change.
I'm carrying out reconstructions, with auto-calibration and both radial and tangential distortion within the adjustment. I output the ppm files, but the tangential distortion coordinates no longer seem to be reporting as non-zero (or maybe they are not being adjusted?). I have previous ppm files in which tangential distortion was being reported fine. The ini file is below.... I'm hoping we can get the parameters to be written again?
Many thanks,
Mike

[CameraCalibration]
FocalX=4444.00
FocalY=4444.00
PrincipalPtX=2000
PrincipalPtY=1500
RadialDistK1=0
RadialDistK2=0
RadialDistK3=0
RadialDistP1=0
RadialDistP2=0
Skew=0
[Cuda]
Device=0
[Export]
Bundlerv03=false
CmvsPmvs2=false
Ogrettha=false
PlyTxt=true
SingleTxtPPMs=true
TextMatlab=false
[Samantha]
AdjustIntrinsic=true
AdjustRadialDistortion=true
AdjustTangentialDistortion=true
ANNApproximationThreshold=0.0
ANNNumberOfFeatures=500
ANNNumberOfNeighbors=8
CheckUpdates=true
ConservativeBa=true
FinalMaxReprojectionError=1.0
FinalMinimumTrackLength=3
ForceSequentialPipeline=false
ImportKeypointsFromTxt=false
KeypointMeanN=7500
LinkageMinSearchRange=1
LocalBa=true
MaxBAIterations=200
MinimumMatches=10
MinimumTrackLength=3
MultithreadANN=true
NumberOfMatchViews=0
PointRecovery=true
ProjectiveReconstruction=true
ReprojectionThreshold=3.0
StructSequence=0
UseSymmetricMatches=true
[Workspace]
Path=F:\2013 James et al SfM accuracies\Prj

[Roberto]

I've checked and the Samantha parameter 'AdjustTangentialDistortion' has been simply unlinked from the new bundle adjustment functions in the command line version. I'll prepared a fixed version later today or tomorrow.
Thanks
Roberto

[Roberto]

I've uploaded a modified versions at this address:
x64
x86
The fix should work, but I haven't had time to test the new version properly yet.
Let me know if you are having problems.

[jamesm]

Thanks Roberto.
I'm getting the tangential components reported now, but error (as given by image residuals on control point observations) has increased significantly. This could be an issue here though - I will check further and compare with my v1.5 results.
All the best,
Mike