XMALab_2.0.0 runs very slowly

[elska_kaczmarek]

I am helping oversee two undergraduate students who are tracking XMA files for us. They have been tracking since early July using XMALab_1.5.5. They switched over to XMALab_2.0.0_beta in September, and then switched to XMALab_2.0.0 a couple of weeks ago when it was released. One of the students told us at the end of last week, that since the update in Sept, XMA has been running incredibly slowly. We checked with the other student and she said that also noticed it running slower since she installed the beta version.

One student has a PC that is less than a year old, and says that the program runs slow regardless of how many frames are in the file. The other student has a mac (Catalina Version 10.15.1) with 10GB of free space out of 120.

Both say that opening XMALab is fine, but tracking is slow. One student specified that moving between frames is fine, but when it is actively automatic tracking it is slow.

[Benjamin Knorlein]

Hi Elska,

I tested xmalab 2.0.0 on 3 different machines now and on all of them i cannot see a problem with speed.

For now i suspect that the issue for this lies in issue #237 as well. If the error message appears xmalab is in a bad state and gets really slow. And it might also be that the problem appears when loading the file but does not show the dialog. Can the student repeat the following test:

1. start xmalab and set in the settings dialog for tracking:

   1. Disable computation of rigid bodies with less than 3 2d correspondances to ON
   2. Optimize Rigid Body transformations based on the detected 2d positions to OFF

2. Then close and restart xmalab and try again with the file and speed test

If the speed is still slow can the student upload the xma-file and the video sequences so that i can check if it is a problem which only occurs with this data and not with mine.

Thanks,

Ben

‌

[elska_kaczmarek]

Hi Ben,

Thanks for testing on different machines and trying to figure this out!

I will have the students try this out and report back. As a heads up, they are in the midst of finals, so it may take a few days for them to get to it.

If it is still slow, then we’ll send you the xma file and image sequences.

Best,

Elska

[elska_kaczmarek]

Hi Ben, 

Sorry for the long delay in replying about further testing of 2.0.0. 

I tested a lab computer and I had three students test 1.5.5 and 2.0.0 with the settings you specified above (Disable computation of rigid bodies with less than 3 2d correspondences to ON; Optimize Rigid Body transformations based on the detected 2d positions to OFF; restart xmalab and do a speed test).

‌

For Ari, Paola, and the lab computer, 2.0.0 is slower than 1.5.5 but not by that much, and changing the tracking settings doesn't make a difference. 

For Noah, changing the tracking settings makes a big difference and mostly closes the gap between 2.0.0 and 1.5.5. 

‌

Noah's tracking is generally slower, maybe there is something specific to his computer that is problematic? He has also recently complained that 1.5.5 keeps crashing when he opens it and that he has to reinstall it in order to get it to open. I'm not sure how to help him with that. 

‌

Overall, it seems like there is less of a problem with 2.0.0 and more of a problem with Noah's computer, though changing the tracking settings certainly does help. 

‌

Here are the numbers:

Noah: (PC, less than a year old)

‌

Lab Comp, remote connection through TeamViewer: (Mac – Mojave 10.14.6)

‌

Ari/Paola: (Mac 10.13.6 / Mac 10.15.1 with 10GB free out of 120GB)

Best,

Elska

[Benjamin Knorlein]

Hi Elsa. Thanks. That is good to know. I will fix the problem with the other issue and will then close both tickets.

Best,

Ben

[Benjamin Knörlein]

fixes #237 and fixes #238 : Changed solvePnP to use EPNP and extended error message with a warning to restart xmalab.

→ <<cset 3e6c12b4deac>>