[tool] [tutorial] Free AI Person Detection for Blue Iris

[astroshare]

Is there a reason you're posting the images using the post tab rather than sending them using the Alert tab? Maybe that is the cause, as I don't see anything else especially wrong with your config. You might also want to remove the "trigger" statement from the Trigger URL because you don't want to re-trigger the camera (or if it is already in a triggered state you might now want to resend a trigger command).

I was just following someone's instructions. I made the changes to use the Alerts tab instead of Post tab and also removed the "trigger" element from the Trigger URL and it seems to be working ok. I'll do more testing tomorrow.

One more question, do I still need a "Cancel URL", or is that not necessary?

Thanks!

 

[Scoobs72]

I was just following someone's instructions. I made the changes to use the Alerts tab instead of Post tab and also removed the "trigger" element from the Trigger URL and it seems to be working ok. I'll do more testing tomorrow.

One more question, do I still need a "Cancel URL", or is that not necessary?

Thanks!

I'm still experimenting myself with the 'Cancel URL' but it seems that if I use Cancel URL (in a set-up where I'm permanently recording 24x7 with motion detection jpeg snapshots) then for any false alerts (i.e. no object of interest detected) the Cancel URL results in the Alert image being removed from the Alert list in Blue Iris. This results in a 'clean' alert list with only confirmed detections in BI, but it also means you need to review the history in AI Tool to see if any valid detections have been missed by Deepstack, rather than in BI itself. For me this is the best approach. But as I say, I'm still experimenting.

 

[joshwah]

This results in a 'clean' alert list with only confirmed detections in BI, but it also means you need to review the history in AI Tool to see if any valid detections have been missed by Deepstack, rather than in BI itself.

You can view them via blue iris... just click on "Cancelled alerts".. they will appear there.

In

 

[austwhite]

I run Deepstack in an auto-restarting Docker container in WSL2. I do this using Task Scheduler as per the answer in this thread.

I have nothing against VMs. I use them for running PiHole, Home Assistant, my NAS, my system monitor, etc. But why do it if you don't have to?

I gave up on the Windows version of Deepstack. I just couldn't get it to run right and it was slow compared to the ubuntu VM I had been using previously.
WSL2 is basically a virtual machine running Linux under Windows and I didn't end up going that route.

I ended up running Deepstack as a docker container on HassOS along side Home Assistant. I installed the Portainer add-in in Home Assistant and pulled the docker image of Deepstack. This way Deepstack loads with Home Assistant.
So I ended up with Blue Iris and AItool running on the host OS and using a Type2 Hypervisor to run Home Assistant on the same machine, Home Assistant runs 24/7 so it was going to be there regardless of which approach I took.
BI is snappier this way, but I still ended up virtualising HA and Deepstack.

 

[joshwah]

Just FYI, i've removed the 'TRIGGER" command in my trigger url. It is working heaps better for me with less processing requirements and more accurate results.

I am using this as my trigger URL:

http://localhost:81/admin?camera=[camera]&user=[Username]&pw=[Password]&flagalert=1&memo=[summary]&jpeg=[ImagePathEscaped]

 

[seth-feinberg]

Personally, I don't actually find the 24/7 low-res recording of much use and I am moving away from it.
You can run all this without cloning cameras and on one camera using version 1.67 or the latest pre-release from VorlonCD. Without a clone, you still save all the motion, but it will only flag recordings with detected objects/persons. If you are okay with this and have the storage space, there is a description how to do it on the first page of this thread using AITool 1.67 or higher. It's pretty simple. You just setup motion recording as normal in Blue Iris, but make it quite sensitive. Then just add the saving JPEG every few seconds to the same camera you record with so AITool has something to send to Deepstack so recordings can be flagged.

As I only want to record the events specifically flagged by AItools/Deepstack and not every motion event, I will still use a cloned camera setup and hide the clones.
The advantage of actual cloned camera over Rob's (The Hookup) setup. The cloned camera uses no extra bandwidth or CPU time. As long as the clone has identical streams to the main camera it is cloned from, it will use no extra resources.
The low res stream Rob uses doesn't use much in resource, but it all adds up if you have several camera's and the low-res JPEGs can sometimes be less accurate, especially if you have a camera observing a wide area, say a camera that is up on the second floor looking out over a yard.

This is a super helpful response, thank you! What is the process for actually cloning the camera? and not doing Rob/The Hookup's method (creating the 2nd stream of cameras)? I'm powering through the thread but it is slow going, i'm only on page 62/147. I've so far seen a screenshot of a radio button to mark "Master Clone" or something to that affect, but nothing about the process of actually cloning (but I assume it's during setup?). I assume I should backup BI first before deleting the low res cameras (or maybe start from scratch)? any advice would be most welcome.

One question i've seen discussed is the max optimal size of the jpeg to analyze. Someone int he 50's of the thread thought 720p was the optimal size, and thought 4k would take too long to process without any added benefit over 720p, any takes on that?

If I understand your method correctly, You are still creating (and hiding) cloned cams, but they are identical resolution (i.e. in my case, 4k). In the end, blue iris will record 24/7 on the 4k "main" stream and flag all events (what clip size do you set here? 8hr? 1hr? and I assume you save ALL footage for a day?) and then on the cloned stream you will record all these motion events for "longer" term storage. is that right?

 

[Sireone]

How are you guys handling alerts, specifically email alerts? I would like to get emails of relative images from AI Tools instead of the normal alerts from BI. Additionally it would be nice if the alert images can contain the same AI indications that is shown in AI Tools History.

 

[mayop]

I've noticed the latest version(s) of BI breaks flagged and confirmed tagging. I've tried it manually in a browser to verify it wasn't AITool and it still didn't work (as of 5.3.7.10).

I rolled it back to 5.3.7.5 and it fixed the issue.

I'm also going to try the confirmed alert tag which will put the Sentry "S" logo in the thumbnails. This might be good for those who don't want to flag verified clips.

/admin?camera=x&flagalert=x&memo=text&jpeg=path
x = 0 mark the most recent alert as cancelled (if not previously confirmed).
x = 1 mark the most recent alert as flagged.
x = 2 mark the most recent alert as confirmed.
x = 3 mark the most recent alert as flagged and confirmed.
x = -1 reset the flagged, confirmed, and cancelled states

 

[Nierka]

I've noticed the latest version(s) of BI breaks flagged and confirmed tagging. I've tried it manually in a browser to verify it wasn't AITool and it still didn't work (as of 5.3.7.10).

I rolled it back to 5.3.7.5 and it fixed the issue.

I'm also going to try the confirmed alert tag which will put the Sentry "S" logo in the thumbnails. This might be good for those who don't want to flag verified clips.

/admin?camera=x&flagalert=x&memo=text&jpeg=path
x = 0 mark the most recent alert as cancelled (if not previously confirmed).
x = 1 mark the most recent alert as flagged.
x = 2 mark the most recent alert as confirmed.
x = 3 mark the most recent alert as flagged and confirmed.
x = -1 reset the flagged, confirmed, and cancelled states

Yes, I can confirm the latest BI broke &flagalert. It drove me crazy, I thought I messed things up.

 

[mayop]

Seems the documentation added a variable called flagclip which seems to flag it properly when I trigger the camera manually and run the url twice in a broswer. Will test to see if it works when BI detects someone.

http://127.0.0.1:81/admin?camera=FrontDoor&flagalert=3&memo=testmessage&user=user&pw=password&flagclip

http://127.0.0.1:81/admin?camera=[camera]&flagalert=3&memo=[summary]&user=[Username]&pw=[Password]&flagclip

Update:
Seems it stopped working again. Must be something broken with the newer updates. Reverted back to 5.3.7.5.

 

[austwhite]

I was scratching my head over why certain sections of video would seem to freeze or skip during the recordings.
I thought it was processing issues, but CPU usage is less than 20%. Deepstack and AITools processing time was less than a second and no repeat triggers being sent.
Then I found this setting in Other under the "Viewer" section called "Skip dead-air during timeline playback".
I didn't really think it would solve it but I thought what the hell and unticked it. All the sudden, no more skipped parts in clip playback.
Not sure if it was the issue, but I am not changing things again now it is running perfectly smoothly.

B.

 

[austwhite]

This is a super helpful response, thank you! What is the process for actually cloning the camera? and not doing Rob/The Hookup's method (creating the 2nd stream of cameras)? I'm powering through the thread but it is slow going, i'm only on page 62/147. I've so far seen a screenshot of a radio button to mark "Master Clone" or something to that affect, but nothing about the process of actually cloning (but I assume it's during setup?). I assume I should backup BI first before deleting the low res cameras (or maybe start from scratch)? any advice would be most welcome.

One question i've seen discussed is the max optimal size of the jpeg to analyze. Someone int he 50's of the thread thought 720p was the optimal size, and thought 4k would take too long to process without any added benefit over 720p, any takes on that?

If I understand your method correctly, You are still creating (and hiding) cloned cams, but they are identical resolution (i.e. in my case, 4k). In the end, blue iris will record 24/7 on the 4k "main" stream and flag all events (what clip size do you set here? 8hr? 1hr? and I assume you save ALL footage for a day?) and then on the cloned stream you will record all these motion events for "longer" term storage. is that right?

Cloning a camera is just adding a new camera and chosing the copy from existing camera option. If the clone has identical streams to the original, it doesn't use extra bandwidth or resources. You then would modify the trigger section of the clone, identical to Rob's method, but chosing never record and only using it to take snapshots.

I use 640x480 resolution JPEGS with a camera that is on the third floor and it picks up people on the ground with about 98% accuracy. I would not use anything above 720p for JPEG resolution unless you want to do full facial recognition as it takes Deepstack longer to process higher resolution images.

If you do not need 24/7 recording, then make sure your clone camera is set to have the Video option not ticked on the Record tab as it will only be used to take JPEG images. So basically there are only 2 differences in the setup over Rob's 24/7 recording setup. 1. Clone has identical feeds to original, rather than only being a substream. 2. Video recording is turned off.
That way the clone only records JPEG's and not video. All video will be recorded on the original camera only when AITools sends a trigger command.

If you want 24/7 recording, you will likely want a low resolution recording stream unless you have huge amounts of storage space to save 4K video or you can tolerate only storing for a very short time and lowering the lifespan of your recording media.

Edit: Just as note, you can hide the cloned cameras and they will still act as triggers and record jpegs. In camera settings on General, click Hidden (make sure enabled remains ticked)

 

[Nierka]

Dopes resi

Cloning a camera is just adding a new camera and chosing the copy from existing camera option. If the clone has identical streams to the original, it doesn't use extra bandwidth or resources. You then would modify the trigger section of the clone, identical to Rob's method, but chosing never record and only using it to take snapshots.

I use 640x480 resolution JPEGS with a camera that is on the third floor and it picks up people on the ground with about 98% accuracy. I would not use anything above 720p for JPEG resolution unless you want to do full facial recognition as it takes Deepstack longer to process higher resolution images.

If you do not need 24/7 recording, then make sure your clone camera is set to have the Video option not ticked on the Record tab as it will only be used to take JPEG images. So basically there are only 2 differences in the setup over Rob's 24/7 recording setup. 1. Clone has identical feeds to original, rather than only being a substream. 2. Video recording is turned off.
That way the clone only records JPEG's and not video. All video will be recorded on the original camera only when AITools sends a trigger command.

If you want 24/7 recording, you will likely want a low resolution recording stream unless you have huge amounts of storage space to save 4K video or you can tolerate only storing for a very short time and lowering the lifespan of your recording media.

Edit: Just as note, you can hide the cloned cameras and they will still act as triggers and record jpegs. In camera settings on General, click Hidden (make sure enabled remains ticked)

Does resize jpeg snapshots is less CPU intensive than DeepStack processing larger images?

 

[austwhite]

Dopes resi

Does resize jpeg snapshots is less CPU intensive than DeepStack processing larger images?

Deepstack uses less CPU time and processes a lot quicker with the smaller images compared to the time Blue Iris uses to resize the image.
Big time saving if you can send Deepstack the resized image and it works out okay for your situation

 

[ZAPme]

Sorry, just catching up after a few days away.

[GIN] 2020/12/29 - 16:26:52 | 200 |    128.7103ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:26:56 | 200 |    144.0097ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:26:59 | 200 |    139.4709ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:27:03 | 200 |    142.1914ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:27:06 | 200 |    127.6803ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:27:10 | 200 |    127.9357ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:27:13 | 200 |    125.9889ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:27:17 | 200 |    124.4378ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:27:20 | 200 |    129.2105ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:27:24 | 200 |    133.2759ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2020/12/29 - 16:27:27 | 200 |    145.9042ms |      172.17.0.1 | POST     /v1/vision/detection

That's on HIGH on a Deepstack GPU running in Docker on WSL2. This is the other instance that runs:

[GIN] 2021/01/01 - 06:27:03 | 200 |    117.6835ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:27:14 | 200 |    135.3589ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:27:17 | 200 |    125.6809ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:27:19 | 200 |    120.2531ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:27:22 | 200 |      119.41ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:30:36 | 200 |    135.4541ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:30:39 | 200 |    132.1701ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:30:41 | 200 |    146.6058ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:30:44 | 200 |    107.0353ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:30:59 | 200 |    136.0961ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:31:02 | 200 |     118.547ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:31:04 | 200 |    105.7385ms |      172.17.0.1 | POST     /v1/vision/detection
[GIN] 2021/01/01 - 06:31:07 | 200 |    107.7303ms |      172.17.0.1 | POST     /v1/vision/detection

I've just noticed nothing logger for a few days. That's not right...

The other question, I get BI to save the JPEGs at 10% quality and 1280x720 from a cloned HD stream with a motion trigger that never records video.

Happy New Year! Thank you for your stats. I had thought the GPU version would be about 10x faster than the CPU version. I may try that cloned HD stream for the jpegs as well.

 

[joshwah]

Anyone having issues with flagging alerts & confirmed etc. not showing up ? Running blue iris 5.3.7.11

 

[mayop]

Anyone having issues with flagging alerts & confirmed etc. not showing up ? Running blue iris 5.3.7.11

Yes it started around the Jan 2 update.

[tool] [tutorial] Free AI Person Detection for Blue Iris

Is there a reason you're posting the images using the post tab rather than sending them using the Alert tab? Maybe that is the cause, as I don't see anything else especially wrong with your config. You might also want to remove the "trigger" statement from the Trigger URL because you don't...

ipcamtalk.com

I rolled back to 5.3.7.5 and they work fine.

 

[kosh42efg]

Happy New Year! Thank you for your stats. I had thought the GPU version would be about 10x faster than the CPU version. I may try that cloned HD stream for the jpegs as well.

My GPU is a low end 1030 Nvidia

 

[astroshare]

I got one flagged alert this morning and then they stopped, even though I can see in AI TOOL that it has positive detections. Is something wrong with my setup in AI TOOL?



EDIT: I've rolled back to 5.3.7.5 and will report back if it's working again

 

[ZAPme]

My GPU is a low end 1030 Nvidia

I have an old GTX660ti with 1300 cuda cores in my Blue Iris server. The 1030 has only 384. I suppose it would make a difference.