Interestingly the Darktable jpgs have less information that the original ones:
-rwx------ 1 stefan stefan 15270635 Jun 8 12:30 Orig Jpeg/DSC_8317.JPG
vs
-rwxrwxr-x 1 stefan stefan 3777924 Jun 16 08:30 Darktable RAW2JPEG/DSC_8317.jpg
Therefore, the control point detector finds less connections between the images and they are of lesser quality. From your ptos:
OrigJpg.pto:
Images are connected by 811 control points, Mean error after optimisation: 3.2 pixels, max 11.1. Good fit
Darktable Jpegs - Fail.pto:
Images are connected by 617 control points, Mean error after optimisation: 41.0 pixels, max 3945.2. Very bad fit
As a result, this project has failed to stitch with the dreaded "Excessive overlap detected" error message.
I have attempted a stitch with hugin 2014.0 RC2. After deleting the outlier points (distance > 30) and re-optimizing for "Positions and View", I got 495 control points, a mean error of 7.3 and max distance of 31.9 which is still a bad fit. However, the project did stitch (see http://hugin.stepet.ch).
To summarize: I can see no specific problem with jpgs from Darktable and Hugin. But your Darktable images seem to miss some information which influences the control point detector negatively. Maybe you need to add some sharpening? Some contrast?
Hi Roland
Interestingly the Darktable jpgs have less information that the original ones:
-rwx------ 1 stefan stefan 15270635 Jun 8 12:30 Orig Jpeg/DSC_8317.JPG DSC_8317. jpg
vs
-rwxrwxr-x 1 stefan stefan 3777924 Jun 16 08:30 Darktable RAW2JPEG/
Therefore, the control point detector finds less connections between the images and they are of lesser quality. From your ptos:
OrigJpg.pto:
Images are connected by 811 control points, Mean error after optimisation: 3.2 pixels, max 11.1. Good fit
Darktable Jpegs - Fail.pto:
Images are connected by 617 control points, Mean error after optimisation: 41.0 pixels, max 3945.2. Very bad fit
As a result, this project has failed to stitch with the dreaded "Excessive overlap detected" error message.
I have attempted a stitch with hugin 2014.0 RC2. After deleting the outlier points (distance > 30) and re-optimizing for "Positions and View", I got 495 control points, a mean error of 7.3 and max distance of 31.9 which is still a bad fit. However, the project did stitch (see http:// hugin.stepet. ch).
To summarize: I can see no specific problem with jpgs from Darktable and Hugin. But your Darktable images seem to miss some information which influences the control point detector negatively. Maybe you need to add some sharpening? Some contrast?
With kind regards
Stefan Peter