Hello,

My current setup uses 8 cameras connected to two PCs, 4 per PC and I use distributed recording obviously.

The problem is, when iPi merges the files at the end, the resulting file has offsets between the master and the slave file. Now this can happen in more ways:

- both the master and the slave file have the same number of frames, yet the resulting file has lesser frames, even with manual merging;
- master file has a no. of frames, the slave a different one, can be 3-4 frames or more, up to 30-40 frames and becomes noticeable for the eye which is how I found out about this issue. In this case the resulting file usually has a completely different no. of frames from both the slave and the master;

Now this is a big issue, as it affects tracking and also calibration. On big gaps you can see limbs just going ahead of the movement by a lot, probably somewhere in between the positions in the two sets of footages while on small gaps a lot of unusual jitter appears.

A quick fix is to edit the master and slave files in the editor by cutting the first say 40 frames in each. This way I can manually merge the files and get the same number of frames in the resulting file.

But this is safe only on master/slave combinations that have the same number of frames, if they are different from where should I know how much to cut? And also, even on same no. of frames files I can't be sure those are the same periods of capture....

It is obviously a synchronization issue and also I believe a mismatch of the timestamps in the files ... or some sort, I am not versed in such matters.

But it is an important issue as it is not rare, last session I recorded 28 files and almost all of them had this problem, some were completely unusable and some had more jitter than normal.

This should be fixed, rather quickly.....

Tnx

Let me give a practical example, a calibration:

slave0 file: 4356 frames
master file: 4314 frames
resulting file: 4312frames

running calibration on the slave0 file (with only four cameras obviously) gives me the following results:

cam1: 0,32% misdetects
0% initial marker misdetection
0% after calibration marker misdetection

cam2 : 0,32% misdetects
0% initial marker misdetection
0% after calibration marker misdetection

cam3 : 1% misdetects
0% initial marker misdetection
1% after calibration marker misdetection

cam4 : 0,32% misdetects
0% initial marker misdetection
0% after calibration marker misdetection

calibration on master file :

cam1 : 0,05% misdetects
0% initial marker misdetection
0% after calibration marker misdetection

cam2 : 0,05% misdetects
0% initial marker misdetection
0% after calibration marker misdetection

cam3 : 0,05% misdetects
0% initial marker misdetection
0% after calibration marker misdetection

cam4 : 0,07% misdetects
1% initial marker misdetection
0% after calibration marker misdetection

Now, in the merged file that iPi creates (the 4312 frames one), I have the same initial misdetections per camera but the misdetection after calibration are all to 4% and this 4% was added also to the final misdetects percentages for each camera, so it is obvious to me that when merging, an offset appeared between the two sets of 4 cameras which led to the after misdetection errors.

When I tried cutting some frames from the beginning of the two files, to "normalize" them I got:

slave0 adjusted: 4310
master adjusted: 4310
resulting file :4310

But the misdetects after calibration all jumped to over 23% for each camera with an after calibration misdetect per camera of same 23 %. And this supports my idea in the previous post that I cannot be sure if the misalignment happens at the beginning of the file, the end... or both or if there is a special method that iPi uses to merge the files. I mean the initial difference between the master/slave files was 42 frames and the resulting file had a 2 frames gap from the master file and 44 from the slave and when I cut them and merged them manually, although the frame number was the same somehow I got a huge initial misdetect error.

By no means is this a bad calibration, especially for an 8 cam setup but knowing that I got an 0,28% misdetects per camera, it's really a shame that this offsetting error prevents me to get such good results every time, not too speak of actual footage were in some cases you can actually see the offsets between the two groups of 4 cameras (half o second or more on a 60 fps recording).

A little follow up.
It seems the footage shot at 60 fps is the one that suffers the most. A couple of shots that we did at 50 fps are perfect, same no. of frames on master/slave and resulting file, or a one frame difference (+/-) at the most.

As a side note, at 50 fps there is almost no flicker, while at 60 we get a lot of flicker. This is strange, because although we have led illumination, running trough a DC converter/transformer, it seems they are picking the 50 Hz frequency of the power grid. Also, the 50/60 Hz camera filter setting doesn't seem to do a thing.

As a quick fix we decided to use a led attached to the wall, that we light briefly at the beginning of the footage and so we can verify the synchronization on the two files. Also this is helpful for us to resynchronize them perfectly with the viewer.