5.5.2 - November 3, 2021

[fenderman]

5.5.2 - November 3, 2021
Export camera status in CSV format.
Status Alerts now offer the ability to fire on both a timed basis and more quickly for
important status updates.
TLS/SSL is now supported for MQTT connections, with the requirement that valid
certificates are provided in PEM format.
When DeepStack is auto-started it is added to a Windows “job” object so that if Blue Iris
exits unexpectedly, DeepStack is closed as well, preventing a system resource leak.

 

[Dave Lonsdale]

Hello fenderman, could there be an easy explanation why when updating from 5.5.1.20 to 5.5.2.0 I no longer get the additional images for Deepstack to choose from? I made no other changes and have "banana" in the cancel box to make sure that all images are checked.

 

[Dave Lonsdale]

5.5.2.1 didn't fix it for me:-



Reverted back to 5.5.1.20, works fine:-


None of my cameras select the best image with 5.5.2/5.5.2.1
Does this happen with anybody else's system?

 

[slabbel]

I had the same issue. Rolled back to 5.5.1.20

 

[Il_Pres]

I report the same. To me it seems that a lot of things are bugged in the latest release:
1) no more HI Res JPEG if the corresponding option is enabled in recording tab
2) many many misses in the DS detection due to the fact that motion rectangles do not overlap to detected objects. it seems to me that the motion rectangle is always referred to the initial detection of movement, not related to the analyzed JPEG.
3) additional JPG setting completely ignored.

@Dave Lonsdale please have a look to your motion rectangles (enable "highlight detected motion rectangles"). Do they overlap to the detected objects?

 

[Dave Lonsdale]

I report the same. To me it seems that a lot of things are bugged in the latest release:
1) no more HI Res JPEG if the corresponding option is enabled in recording tab
2) many many misses in the DS detection due to the fact that motion rectangles do not overlap to detected objects. it seems to me that the motion rectangle is always referred to the initial detection of movement, not related to the analyzed JPEG.
3) additional JPG setting completely ignored.

@Dave Lonsdale please have a look to your motion rectangles (enable "highlight detected motion rectangles"). Do they overlap to the detected objects?

Hello Il_Pres
I have exactly the same problem that you identified in your number 2). I don’t get the additional images that I specify for DeepStack and no part of the highlight overlaps the motion rectangle in the two images that remain. The highlight does not attempt to track the motion, perhaps because the alert is cancelled after the second image and this is also when the analysis delay is shorter than the interval specified between the setting for the additional images (if that makes any difference). The problem is not dependent on whether or not “Use main stream if available“ is checked.

This happens on all my cameras unless I roll back to 5.5.1.20. Surely others have the same problem? Did you send a message to BI Support?

PS. I’m guessing your not Italian - your english is perfect…

 

[Dave Lonsdale]

Not fixed in 5.5.2.2. On my system, the highlight box stays put and the alert is canceled without the additional images having been analysed.

Is it only slabbel, Il-Pres and me that have this problem?

 

[bqz]

5.5.2.2 is awful

 

[tward392]

I emailing back and forth with support as 5.5.2.2 seems to have stopped DeepStack from working for me completely.

 

[IReallyLikePizza2]

5.5.2.1 didn't fix it for me:-

View attachment 107453

Reverted back to 5.5.1.20, works fine:-

View attachment 107454
None of my cameras select the best image with 5.5.2/5.5.2.1
Does this happen with anybody else's system?

Lovely property!

 

[Il_Pres]

PS. I’m guessing your not Italian - your english is perfect…

That's mighty flattering of you... I've just experienced my English sending bug reports to Ken... joking

To be back on topic, for now I'm back to AITools. I've got so much more stability and control over the processed images. Will follow the evolution of BI and migrate when it'll be more stable. For my purposes, BI's been unreliable as every release changes the behavior of the DS engine in some way or another.

 

[kc8tmv]

Anyone brave enough to try the 5.5.2.4 yet?

 

[sebastiantombs]

I updated to .4 last night and it seems fine to me on my system. YMMV though.

 

[kc8tmv]

I updated to .4 last night and it seems fine to me on my system. YMMV though.

I tried .1, .2 and .3 just long enough to see that they just did not work. Didn't need to even get into the details, just obvious it was not going to work. I might just hang on a few more releases.
I usually watch to see what you (and a few others) say before diving down the rabbit hole!

Did they ever post the webinar showing the new "features"? If so, where?

 

[sebastiantombs]

No, there's no webinar or even information that is specific to the updates in the "what's new" section either. I sure would like to know what the update was aimed at other than the boiler plate that gets posted but I don't think that's going to happen.

 

[kc8tmv]

No, there's no webinar or even information that is specific to the updates in the "what's new" section either. I sure would like to know what the update was aimed at other than the boiler plate that gets posted but I don't think that's going to happen.

I develop software (old school 4GL, nothing fun like this stuff), but if I didn't have a full set of release notes, it would never make it to the users!

 

[Dave Lonsdale]

I develop software (old school 4GL, nothing fun like this stuff), but if I didn't have a full set of release notes, it would never make it to the users!

I think you guys are being far too picky. Ken Pletzer does a brilliant job of responding to the Beta team feedback (you guys). It's obvious that 5.5.2.4 was to fix problems identified in 5.5.2.x.

I would, nevertheless, like to understand this highlight thing better. Looking at the pictures below, I think the first one is where motion with a person and a highlight is first detected (begin analysis with motion leading image is checked). The second one is the first BI red rectangle trigger - I wonder why the highlight is absent? It seems that the second image in the 'series' at the instant of trigger on a confirmed alert is 'blue' on all my cameras. The third picture is when the highlight has caught up. Why is the highlight overlay needed when BI has already triggered?

Notice that the image selected with the asterisk has the highest percentage confidence - excellent! But it's a good job I have additional images selected for analysis...

 

[Il_Pres]

I think you guys are being far too picky. Ken Pletzer does a brilliant job of responding to the Beta team feedback (you guys). It's obvious that 5.5.2.4 was to fix problems identified in 5.5.2.x.

I would, nevertheless, like to understand this highlight thing better. Looking at the pictures below, I think the first one is where motion with a person and a highlight is first detected (begin analysis with motion leading image is checked). The second one is the first BI red rectangle trigger - I wonder why the highlight is absent? It seems that the second image in the 'series' at the instant of trigger on a confirmed alert is 'blue' on all my cameras. The third picture is when the highlight has caught up. Why is the highlight overlay needed when BI has already triggered?

Notice that the image selected with the asterisk has the highest percentage confidence - excellent! But it's a good job I have additional images selected for analysis...

I confirm the same behaviour on .4: the T-XX's (ie: T-78, T-66.... can be second analysed image or first, if no leading ) have highlight rectangle all wrong (not present or related to leading image motion), cancelling detection. I have some cameras that are set to analyze just one image, with no leading and they're not firing alerts because of the detected object box is not in the same position of the motion highlight on T-XX images. A little more debug step is required here, IMO.

 

[Il_Pres]

Here's proof of what I'm saying. This was a missed alert of a car travelling throug the driveway.

 

[tward392]

Here's proof of what I'm saying. This was a missed alert of a car travelling throug the driveway.

I've been getting the same thing.