Hello! My use case involves collecting an SVO2 ranging between 400-500mb in a farm. This SVO is then uploaded to cloud for downstream processing.
What I found was that uploading a 500mb file in wifi-poor environments like farms leads to repeated upload failures. Through experiments, I found that smaller files have a higher upload success rate.
Suppose a safe chunk size is 50mb, I would divide my 500mb file into 10 chunks. I identified the ranges of frames that belong to each chunk, then proceeded to cut the SVO into 10 chunks using a command like this:
“ZED_SVO_Editor -cut {svo_path} -s {start} -e {end} {split_svo_path}”
The resulting chunks ought to be close to 50mb, but this is not what I observed. On the contrary, the resulting chunks are significantly less than 50mb.
These are some logs from a specific SVO File, with target chunk size of 50mb:
SVO File is 568.2300071716309 mb
Number of frames in SVO file: 3097 → gotten from sl.Camera().get_svo_number_of_frames()
Number of chunks: 11
Frames per chunk: 281
These are the start and end frame indexes of each chunk (11 chunks in total):
[60, 340]
[341, 621]
[622, 902]
[903, 1183]
[1184, 1464]
[1465, 1745]
[1746, 2026]
[2027, 2307]
[2308, 2588]
[2589, 2869]
[2870, 3096]
These are the sizes of the resulting chunks:
recording_0.svo2 Size: 28.671786308288574 mbs
recording_1.svo2 Size: 27.706414222717285 mbs
recording_2.svo2 Size: 27.863109588623047 mbs
recording_3.svo2 Size: 28.039546966552734 mbs
recording_4.svo2 Size: 27.67126750946045 mbs
recording_5.svo2 Size: 28.16982650756836 mbs
recording_6.svo2 Size: 28.056971549987793 mbs
recording_7.svo2 Size: 28.13659954071045 mbs
recording_8.svo2 Size: 28.23575496673584 mbs
recording_9.svo2 Size: 28.133583068847656 mbs
recording_10.svo2 Size: 22.00721549987793 mbs
I then recombine all of them again to compare with the original. The recombined size is 307mb. The original was 568mb.
Why is there this discrepancy??
ZED SDK VERSION 4.2.3