search

wilsonkl / SfM_Init

code for solving global structure from motion problem in the ECCV'14 paper "Robust Global Translations with 1DSfM"
Other
172 stars 60 forks source link

aberations in trans_problem_solution.txt #8

Closed alvaro562003 closed 7 years ago

alvaro562003 commented 7 years ago

Hello Kyle,

I am still integrating SFM_init in a classic matlab SFM project.

Sometimes in the file trans_problem_solution.txt there are some aberations. For example at line 21 below 20 -0.133264708712 0.992133578918 1.66608261415 21 -368554.531735 103353.696908 90712.47405 22 0.161763972331 -0.10153434714 0.122690909483

All other files "seems" good.

Do you have ever encountered this type of issue ? And What should i do to correct this error ? (a subtle file format rule not respected ?)

Thank you for your help

MotionAveraging.zip

wilsonkl commented 7 years ago

Hi Alvaro,

Can you explain the problem in more detail? This looks like valid formatting to me. Sometimes parts of translations problems are ill-conditioned. Small noise can result in point locations that are very far away from the rest of the model. That may well me the correct solution, or something close to it.

On Mon, Jan 23, 2017 at 8:11 AM, alvaro562003 notifications@github.com wrote:

Hello Kyle,

I am still integrating SFM_init in a classic matlab SFM project.

Sometimes in the file trans_problem_solution.txt there are some aberations. For example at line 21 below 20 -0.133264708712 0.992133578918 1.66608261415 21 -368554.531735 103353.696908 90712.47405 22 0.161763972331 -0.10153434714 0.122690909483

All other files "seems" good.

Do you have ever encountered this type of issue ? And What should i do to correct this error ? (a subtle file format rule not respected ?)

Thank you for your help

MotionAveraging.zip https://github.com/wilsonkl/SfM_Init/files/723634/MotionAveraging.zip

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/wilsonkl/SfM_Init/issues/8, or mute the thread https://github.com/notifications/unsubscribe-auth/ACmoVkb_JmlB0MhjKeJ3WKDrGhnNtN2Sks5rVKb-gaJpZM4LrBcy .

alvaro562003 commented 7 years ago

Hello Kylen Indeed there were noise in my data. After correction there were no more abberations. Thank you for having share your experience.

By the way, what is the minimum images needed to use SFM_init ?

wilsonkl commented 7 years ago

A problem could be well-posed with as few as three images, if they were not co-linear. With using scene points, the code could run with just two images. I'm not sure that this would be useful for anything, but it is definitely the lower bound.

On Wed, Feb 1, 2017 at 12:08 AM, alvaro562003 notifications@github.com wrote:

Hello Kylen Indeed there were noise in my data. After correction there were no more abberations. Thank you for having share your experience.

By the way, what is the minimum images needed to use SFM_init ?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/wilsonkl/SfM_Init/issues/8#issuecomment-276573822, or mute the thread https://github.com/notifications/unsubscribe-auth/ACmoVhdR2WWJRjQJViqKaSZy74DzdkGmks5rYBM1gaJpZM4LrBcy .