Computing Covariance - How to improve performance

199 views
Skip to first unread message

Alan Buchanan

unread,
Jul 11, 2017, 8:47:44 AM7/11/17
to Ceres Solver
Hi Sameer,

I am computing the covariance for the interior and exterior orientation from a reasonably large dataset (750 high resolution 8100x4530 pixels).

The performance is very slow whereas the rest of the adjustment is reasonably fast. It is still crunching numbers and has been going for over an hour.

I am using openMVG with the default Ceres install on a windows machine (12GB ram and a decent graphics card).

I am using the following options.

ceres::Covariance::Options options;
options.num_threads = 7;                           (I changed to 7 from the default of 1)
options.algorithm_type = ceres::EIGEN_SPARSE_QR;
ceres::Covariance covariance(options);

The rest is as per the example in the documentation

Are there any tricks to speeding this up or is it merely the fact we are trying to invert a very large matrix and have t live with it.

Regards,

Alan

Sameer Agarwal

unread,
Jul 11, 2017, 9:29:55 AM7/11/17
to Ceres Solver
options.algorithm_type = SUITE_SPARSE_QR (You will need SuiteSparseQR installed for this to work) is your best bet to do this right now.
Sameer


--
You received this message because you are subscribed to the Google Groups "Ceres Solver" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ceres-solver...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ceres-solver/5cd9f6a0-7fc6-4406-b4ba-dcbb4d61af04%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Alan Buchanan

unread,
Jul 11, 2017, 10:27:24 AM7/11/17
to ceres-...@googlegroups.com
Thanks Same error I will have look at downloading and installing this



On 11 Jul 2017 21:29, "'Sameer Agarwal' via Ceres Solver" <ceres-...@googlegroups.com> wrote:
options.algorithm_type = SUITE_SPARSE_QR (You will need SuiteSparseQR installed for this to work) is your best bet to do this right now.
Sameer


On Tue, Jul 11, 2017 at 5:47 AM Alan Buchanan <abuc...@gmail.com> wrote:
Hi Sameer,

I am computing the covariance for the interior and exterior orientation from a reasonably large dataset (750 high resolution 8100x4530 pixels).

The performance is very slow whereas the rest of the adjustment is reasonably fast. It is still crunching numbers and has been going for over an hour.

I am using openMVG with the default Ceres install on a windows machine (12GB ram and a decent graphics card).

I am using the following options.

ceres::Covariance::Options options;
options.num_threads = 7;                           (I changed to 7 from the default of 1)
options.algorithm_type = ceres::EIGEN_SPARSE_QR;
ceres::Covariance covariance(options);

The rest is as per the example in the documentation

Are there any tricks to speeding this up or is it merely the fact we are trying to invert a very large matrix and have t live with it.

Regards,

Alan

--
You received this message because you are subscribed to the Google Groups "Ceres Solver" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ceres-solver+unsubscribe@googlegroups.com.

--
You received this message because you are subscribed to a topic in the Google Groups "Ceres Solver" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/ceres-solver/wJXIjU00Mdw/unsubscribe.
To unsubscribe from this group and all its topics, send an email to ceres-solver+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ceres-solver/CABqdRUBC45OGh%2BD5-a-41md8DpFGBK3Mmf5_CmVf6DuBntNQvA%40mail.gmail.com.

Alan Buchanan

unread,
Jul 12, 2017, 6:53:24 AM7/12/17
to Ceres Solver
Just hold off on my last question.

I noticed that although I enable suite sparse in Cmake it did not generate a project with it. Apparantly I have to also include LAPACK.

I will let you know later if I still have problems

Alan Buchanan

unread,
Jul 12, 2017, 8:19:56 AM7/12/17
to Ceres Solver
Sameer,

I have spent a lot of time trying install ceres with suit sparse on windows. Looking on google it appears that other people are having issues with installing suit sparse and Lapack.

I have ubuntu linux box at home that I think I will bite the bullet and install on there as it looks much easier.

So I will close off this thread until after I test on a different platform


On Tuesday, 11 July 2017 20:47:44 UTC+8, Alan Buchanan wrote:
Reply all
Reply to author
Forward
0 new messages