Here is a issue about cpfind performance in version 2017.0.0. I use this command
cpfind --prealigned -o {origin_pto} {final_pto}
to update control points in pto based on six fisheye images.
The cpfind command costs about 90s in hugin version 2017.0.0, but only costs about 30s in hugin version 2016.2.0
I see the changed about cpfind in 2017 release notes, should I change my command arguments or maybe is it a performance bug?
Thank you for your help.
The cpfind command costs about 90s in hugin version 2017.0.0, but only costs about 30s in hugin version 2016.2.0I see the changed about cpfind in 2017 release notes, should I change my command arguments or maybe is it a performance bug?
add-apt-repository -y ppa:hugin/hugin-builds
apt-get update
apt-get install -y hugin
Hugin's cpfind 2017.0.0.eac5e8cc546e
based on Pan-o-matic by Anael Orlinski
Input file : test1/8bd82ed4bd9df7159d006cfa5e6ea0db.pto
Output file : default.pto
Number of threads : 2
Input image options
Downscale to half-size : yes
Sieve 1 Options
Width : 10
Height : 10
Size : 100
==> Maximum keypoints per image : 10000
KDTree Options
Search steps : 200
Second match distance : 0.25
......
--- Write Project output ---Written output to default.ptoDetection took 97.044 seconds
my os is ubuntu 14.04I install hugin by apt-getadd-apt-repository -y ppa:hugin/hugin-builds
apt-get updateapt-get install -y huginthe "cpfind -v demo.ptoi" showsHugin's cpfind 2017.0.0.eac5e8cc546ebased on Pan-o-matic by Anael OrlinskiInput file : test1/8bd82ed4bd9df7159d006cfa5e6ea0db.pto
Output file : default.pto
Number of threads : 2
Also your pto file contains some strange settings: a FOV of 357.4 is very big, also for a fisheye. Also the lens distortion parameter a very high (a1.2 b-2.5 c0.46), probably to tweak the wrong fov.