Responses to user support requests. Weekly overview. Part 18.
Q: I just noticed Xeoma server also decided to upgrade itself to latest version on my VM.
A: There’s a -noautoupdate command that stops server from being updated automatically
Q: We need 90% accuracy for video analytics
A: Regarding 90+% accuracy
Besides the algorithms, the accuracy of recognition depends on the location of the camera and external conditions. Let’s consider this on the emotion recognition case. So, how to increase the percentage of emotion recognition?
• You can hang the camera as close as possible to the area in which you need to detect emotions (preferably at a right angle to the face)
• The angle should be such that the face takes up most of the frame
• The lighting should not be very dim or with a lot of highlights (you can use special cameras with the HLC function – High Light Compensation (compensation for bright illumination, often marked: for LPR / ANPR))
• You can also use a long-focus lens for a better view
Q: Can i know, if i will need to do post event investigation with privacy masking on the recording, can i know the proposed camera recording need to store in the server?
A: In order to answer this question, I’ll need to explain a bit of how Xeoma’s archiving works. There are 2 ways to store archives:
1) using the same codec as the camera provided for the original stream (usually it’s H.264 or H.265)
2) using MJPEG
The first method is very efficient in terms of storage space, since the video remains compressed. However, that also means that Xeoma cannot alter anything in the video itself, it has to be stored exactly as is. This means the resulting video files (.mkv in our case) cannot have privacy masking applied to them.
The second method allows Xeoma to apply any changes to the video (including privacy masking), but is much more demanding in terms of storage space.
Current versions of Xeoma can still apply privacy masking for the first method when you are watching the archives via Xeoma’s own archive player (the client side handles that). However, if you take the separate MKV files or perform an export of a video segment, no privacy masking will be present in the result, since this is something the client side cannot affect.
September 19, 2025