bogus (?) warning messages about pixel size when running relion_motion_refine

Issue #16 resolved
Oliver Clarke created an issue

Hi,

When I run relion_motion_refine in training mode, I sometimes get a (non-fatal) error that the movie pixel size and coord pixel size do not match, even though they do (and are listed in the log file as being identical). Explicitly setting all the pixel sizes the same (on the command line, using --coords_angpix, --movie_angpix and --angpix) does not remove the error. Seems harmless, job still runs etc, but it is a little confusing.

Cheers Oli

Comments (6)

  1. Jasenko Zivanov

    We are a little surprised that you would see this message. For it to appear, one number would have to be at least one billionth smaller than the other.

    The reason we are somewhat paranoid about pixel size precision is that it is currently the only way to determine the box size in the original movie. If the pixel size is off by only a quarter of a percent (and it is in some datasets that we have seen), it will almost certainly change the box size in movie pixels by at least one pixel. Then, the reference won't fit the images anymore, and the motion estimation will be less reliable.

  2. Oliver Clarke reporter

    Yes, but as you can see from the messages, the pixel sizes are actually the same. After seeing the messages initially, I used the --angpix and --movie_angpix flags to make sure they were the same, so I'm not sure what is going wrong

  3. Jasenko Zivanov

    They are obviously not the same! :)

    The question is, why do they differ by more than a billionth even though you input the same numbers. There is certainly some technical explanation for this that has to do with IEEE floating point precision.

    Does this only happen during parameter estimation and not during the actual motion fit?

    I'm confident that it's perfectly harmless, though. As long as your pixel sizes are integral multiples of each other, they won't cause any problems.

    We apologize for the confusion!

  4. Oliver Clarke reporter

    I get the same warning during the motion fit... but good to know it is harmless!

    Cheers Oli

  5. Log in to comment