Blue Iris and Deep Stack Fresh Install - 2 Questions

[rjp1267]

Hello I am installing BI and DS on a new pc and following the you tuber "The Hook Up" and have 2 questions regarding the instructions within the video.

1. He sets up hd and sd streams, the sd is for 24 hour recording and sd for hi res capture. After using this setup for a couple of days I am seeing my ssd space filling up quickly with low res 24 hr recordings and questioning why am I doing this? I am installing DS to reduce the false notifications and increase identification accuracy alerting to 80% or better and use BI to automate some functions, 24 hr recording to my way of thinking isnt providing much if any value. Put another way I am content with capturing events I consider important and leave the rest alone. Also if I choose to forgo the 24 hr sd recording I can increase the image quality of my desktop and web interface. Any thoughts, experience, feedback?

2. In AI Tool, which is used to define what you want DS to look for, there is a "Trigger URL" which sends a trigger command to the camera to start recording when an event occurs. The question I have is this...Is the URL a standard format meaning it works with all IP cameras or does it vary by manufacturer? I have Amcrest and Annke, the latter I think I have solved as the guy in the video also uses Annke, but I am still trying to find any info for Amcrest. Ive done searches for "rtsp url trigger" and variations with no luck. Any advice appreciated.

Thank you.

 

[wittaj]

First, don't follow the Hook Up LOL

Second, if you use the function of BI and record "continuous + triggers" then it will record substream until triggered, then switch to mainstream for the trigger. Saves you from having to record two streams.

Third, DeepStack is integrated with BI to make the process simple. Unless you need fine granular level control, AI Tool is overkill for most people. No need for Docker and other stuff with the integrated version.

Simply Download the Deepstack for CPU and install the executable. Then go into the BI Settings and tell it to use DeepStack, and then go into each camera you want to use DeepStack and under Trigger and tell it to use Deepstack. Done... (well then some fine tuning of triggers, etc.), but you can get going quickly. And no need to know the trigger URL and rtsp url trigger nonsense you are dealing with now. If the camera is in BI, it will work.

Suggest just doing it to one camera and learn the nuances first and then add it to other cameras.

Using DeepStack with Windows 10 (CPU and GPU)

docs.deepstack.cc

 

[sebastiantombs]

^^^^ What he said.

Following any advice from "Hook Up" is a guarantee to have a failure.

 

[rjp1267]

First, don't follow the Hook Up LOL

Second, if you use the function of BI and record "continuous + triggers" then it will record substream until triggered, then switch to mainstream for the trigger. Saves you from having to record two streams.

Third, DeepStack is integrated with BI to make the process simple. Unless you need fine granular level control, AI Tool is overkill for most people. No need for Docker and other stuff with the integrated version.

Simply Download the Deepstack for CPU and install the executable. Then go into the BI Settings and tell it to use DeepStack, and then go into each camera you want to use DeepStack and under Trigger and tell it to use Deepstack. Done... (well then some fine tuning of triggers, etc.), but you can get going quickly. And no need to know the trigger URL and rtsp url trigger nonsense you are dealing with now. If the camera is in BI, it will work.

Suggest just doing it to one camera and learn the nuances first and then add it to other cameras.

Using DeepStack with Windows 10 (CPU and GPU)

docs.deepstack.cc

Thank you for your response, being new to the BI and DS world that video hit the mark with everything I am looking to achieve and the process didnt seem so difficult but am glad I checked in here before fully committing. As you may expect I am awash with questions so please bear with me.

1. I have looked at the AI configuration tab within BI and it seems straight forward, but where do you tweak the accuracy/sensitivity so that false positives are not an issue? The reason why I am integrating DS is to ensure that notifications are highly reliable (>80%) (no shadows, rain, cob webs, etc causing a trigger) and that I capture the events that are of importance. I would like to eventually use AI in the future to control what happens when say a recognized/authorized face or license plate is detected like not record, or turn on a light or some other action.

2. It seems that BI can be configured to point to the DS installation location (AI tab) for auto start/stop, just want to confirm that since I have BI set as a windows service that DS will also start when BI starts?

3. For whatever reason the video has you install/run DS through a container, can DS and BI run off the same windows box (no container or VM) without issues?

4. Can I still try out AI Tools using the local windows installation method of DS?

5. Lastly, back to the live stream recording, is this really necessary or can I just configure to capture hi res video when an event is triggered? I dont need to have a 24 hour archive of what the cameras see, only those events I configure. Takes up a great deal of disk space and is of no value to me, so not sure if required for DS to work properly or I can just setup to capture when triggered?

Thank you for your patience and support.

 

[wittaj]

Answers to your questions:

WIthin the camera setting you put down the confidence level you want for the item.

Yes, if you use the integrated DeepStack/BI then DeepStack will start when BI does.

You do not need to run it thru a docker/container/VM if you use the integrated DeepStack/BI. Container/Docker, etc. is only needed if you are using the 3rd party AI Tool type programs. People run this integrated DeepStack/BI on a 4th generation computer just fine.

AI Tools needs the container/docker so it will not work with the Windows version I linked above. Do not trust the hook up video LOL. Next you will say you bought Reolinks based on his recommendation

Substreams do not take that much space. Murphy's law is if you only record on triggers, inevitably something will be missed. As good as these cameras and programs are, stuff can still pass by undetected, and that is where the substream 24/7 helps. Fact of the matter is, if it happened more than two weeks ago and you didn't notice it, the likelihood you would scrub through months of video/triggers is slim to none. BI can be so customized that you can get alerts for the areas of importance to you. I can literally scrub my overnight feeds in minutes in the morning. Usually under a minute. But if you don't want to, that is your choice - but I can assure you that you will miss something and you will not take the time to look at something 3 months ago LOL.

With BI update 5.4.4, a new feature was added that results in significant savings in storage. When the triggered+continuous record mode is used with a dual-streaming camera along with direct-to-disc, the result is a BVR file which will contain the sub-stream continuously recorded, but the main-stream only recorded when the camera is in a triggered state. During main-stream playback, the sub-stream will be upsampled whenever the main-stream is not available.

I tested this with one camera overnight to see how it works.

Previously, I would get about 1 hour at 3.90GB with the camera recording 24/7.

With this new feature, I got 1 hour at 120MB.

That is a tremendous savings!

As always YMMV. Obviously bitrate and FPS come into play, but it is yet another feature that can allow for higher resolutions with less storage needs.

BI stores each event as an alert, so it is really easy to scrub through those and only have to look at the non-triggered video in those instances where the camera didn't trigger.


Regarding facial recognition, keep in mind that most of us have found in home settings that facial recognition is more gimmicky and novelty than anything else. If you have to put in 5 or 10 or 15 or 35 pictures of yourself in system for it to recognize it is you...then you shouldn't expect much.... My success rate was under 5% so I moved on to other things LOL.

It can work in certain situations like a business that requires everyone to stop in front of the camera and the camera is at head height. Outside of that, the percentage of being accurate is probably not going to be super high. Especially someone coming to a front door.

Unless you spend the big bucks that casinos and airports have LOL.

Heck even in ideal situations like a business with the camera at ideal height and optimal lighting it fails....

What do you think of Dahua's "Uyghur Warnings"?

Good luck. Wish you no ill will. I'm not here to argue with people simply to contribute when and where I can. This is also followed by lending my views on current technology and how it impacts all of us. It's undisputed that splitting of the atom changed the world but once again man decided to...

ipcamtalk.com

Regarding a camera for plates (LPR) - keep in mind that this is a camera dedicated to plates and not an overview camera also. It is as much an art as it is a science. If you want to capture plates, you need a camera just for that.

During daytime it may act as an overview, but at night the shutter tends to need to be faster to capture the plate, and the faster the shutter, the darker the image.

Most of us run B/W to take advantage of the reflective properties of a plate which is why we can run a faster shutter to get the plates. Once you decide to keep it in color, the shutter speed slows way down and might not be fast enough to actually capture a plate.

You may have enough light and the car moving slower to capture, but you may also need to decide what is more important - the overview or the plate. Most of us run two cameras.

At night, we have to run a very fast shutter speed (1/2,000) and in B/W with IR and the image will be black. All you will see are head/tail lights and the plate. Some people can get away with color if they have enough street lights, but most of us cannot. Here is a representative sample of plates I get at night of vehicles traveling about 45MPH at 175 feet from my 2MP camera (that is all that is needed for plates):

 

[rjp1267]

Answers to your questions:

WIthin the camera setting you put down the confidence level you want for the item.

Yes, if you use the integrated DeepStack/BI then DeepStack will start when BI does.

You do not need to run it thru a docker/container/VM if you use the integrated DeepStack/BI. Container/Docker, etc. is only needed if you are using the 3rd party AI Tool type programs. People run this integrated DeepStack/BI on a 4th generation computer just fine.

AI Tools needs the container/docker so it will not work with the Windows version I linked above. Do not trust the hook up video LOL. Next you will say you bought Reolinks based on his recommendation

Substreams do not take that much space. Murphy's law is if you only record on triggers, inevitably something will be missed. As good as these cameras and programs are, stuff can still pass by undetected, and that is where the substream 24/7 helps. Fact of the matter is, if it happened more than two weeks ago and you didn't notice it, the likelihood you would scrub through months of video/triggers is slim to none. BI can be so customized that you can get alerts for the areas of importance to you. I can literally scrub my overnight feeds in minutes in the morning. Usually under a minute. But if you don't want to, that is your choice - but I can assure you that you will miss something and you will not take the time to look at something 3 months ago LOL.

With BI update 5.4.4, a new feature was added that results in significant savings in storage. When the triggered+continuous record mode is used with a dual-streaming camera along with direct-to-disc, the result is a BVR file which will contain the sub-stream continuously recorded, but the main-stream only recorded when the camera is in a triggered state. During main-stream playback, the sub-stream will be upsampled whenever the main-stream is not available.

I tested this with one camera overnight to see how it works.

Previously, I would get about 1 hour at 3.90GB with the camera recording 24/7.

With this new feature, I got 1 hour at 120MB.

That is a tremendous savings!

As always YMMV. Obviously bitrate and FPS come into play, but it is yet another feature that can allow for higher resolutions with less storage needs.

BI stores each event as an alert, so it is really easy to scrub through those and only have to look at the non-triggered video in those instances where the camera didn't trigger.


Regarding facial recognition, keep in mind that most of us have found in home settings that facial recognition is more gimmicky and novelty than anything else. If you have to put in 5 or 10 or 15 or 35 pictures of yourself in system for it to recognize it is you...then you shouldn't expect much.... My success rate was under 5% so I moved on to other things LOL.

It can work in certain situations like a business that requires everyone to stop in front of the camera and the camera is at head height. Outside of that, the percentage of being accurate is probably not going to be super high. Especially someone coming to a front door.

Unless you spend the big bucks that casinos and airports have LOL.

Heck even in ideal situations like a business with the camera at ideal height and optimal lighting it fails....

What do you think of Dahua's "Uyghur Warnings"?

Good luck. Wish you no ill will. I'm not here to argue with people simply to contribute when and where I can. This is also followed by lending my views on current technology and how it impacts all of us. It's undisputed that splitting of the atom changed the world but once again man decided to...

ipcamtalk.com

Regarding a camera for plates (LPR) - keep in mind that this is a camera dedicated to plates and not an overview camera also. It is as much an art as it is a science. If you want to capture plates, you need a camera just for that.

During daytime it may act as an overview, but at night the shutter tends to need to be faster to capture the plate, and the faster the shutter, the darker the image.

Most of us run B/W to take advantage of the reflective properties of a plate which is why we can run a faster shutter to get the plates. Once you decide to keep it in color, the shutter speed slows way down and might not be fast enough to actually capture a plate.

You may have enough light and the car moving slower to capture, but you may also need to decide what is more important - the overview or the plate. Most of us run two cameras.

At night, we have to run a very fast shutter speed (1/2,000) and in B/W with IR and the image will be black. All you will see are head/tail lights and the plate. Some people can get away with color if they have enough street lights, but most of us cannot. Here is a representative sample of plates I get at night of vehicles traveling about 45MPH at 175 feet from my 2MP camera (that is all that is needed for plates):

View attachment 111403

Lots of great stuff here. I have taken a step back to decide how best to proceed, I have removed the AI tools, the container and am now deciding if it would be best to uninstall/reinstall BI and work from there. Oddly enough a recommendation from ipcamtalk pointed to a BI/DS install check list and the video it pointed to for BI/DS install was the hoook up guy and this time his directions reflect what has been suggested here. I want to do it right so am walking softly and digesting what has been suggested thus far. Will be back shortly with status.

 

[rjp1267]

Answers to your questions:

WIthin the camera setting you put down the confidence level you want for the item.

Yes, if you use the integrated DeepStack/BI then DeepStack will start when BI does.

You do not need to run it thru a docker/container/VM if you use the integrated DeepStack/BI. Container/Docker, etc. is only needed if you are using the 3rd party AI Tool type programs. People run this integrated DeepStack/BI on a 4th generation computer just fine.

AI Tools needs the container/docker so it will not work with the Windows version I linked above. Do not trust the hook up video LOL. Next you will say you bought Reolinks based on his recommendation

Substreams do not take that much space. Murphy's law is if you only record on triggers, inevitably something will be missed. As good as these cameras and programs are, stuff can still pass by undetected, and that is where the substream 24/7 helps. Fact of the matter is, if it happened more than two weeks ago and you didn't notice it, the likelihood you would scrub through months of video/triggers is slim to none. BI can be so customized that you can get alerts for the areas of importance to you. I can literally scrub my overnight feeds in minutes in the morning. Usually under a minute. But if you don't want to, that is your choice - but I can assure you that you will miss something and you will not take the time to look at something 3 months ago LOL.

With BI update 5.4.4, a new feature was added that results in significant savings in storage. When the triggered+continuous record mode is used with a dual-streaming camera along with direct-to-disc, the result is a BVR file which will contain the sub-stream continuously recorded, but the main-stream only recorded when the camera is in a triggered state. During main-stream playback, the sub-stream will be upsampled whenever the main-stream is not available.

I tested this with one camera overnight to see how it works.

Previously, I would get about 1 hour at 3.90GB with the camera recording 24/7.

With this new feature, I got 1 hour at 120MB.

That is a tremendous savings!

As always YMMV. Obviously bitrate and FPS come into play, but it is yet another feature that can allow for higher resolutions with less storage needs.

BI stores each event as an alert, so it is really easy to scrub through those and only have to look at the non-triggered video in those instances where the camera didn't trigger.


Regarding facial recognition, keep in mind that most of us have found in home settings that facial recognition is more gimmicky and novelty than anything else. If you have to put in 5 or 10 or 15 or 35 pictures of yourself in system for it to recognize it is you...then you shouldn't expect much.... My success rate was under 5% so I moved on to other things LOL.

It can work in certain situations like a business that requires everyone to stop in front of the camera and the camera is at head height. Outside of that, the percentage of being accurate is probably not going to be super high. Especially someone coming to a front door.

Unless you spend the big bucks that casinos and airports have LOL.

Heck even in ideal situations like a business with the camera at ideal height and optimal lighting it fails....

What do you think of Dahua's "Uyghur Warnings"?

Good luck. Wish you no ill will. I'm not here to argue with people simply to contribute when and where I can. This is also followed by lending my views on current technology and how it impacts all of us. It's undisputed that splitting of the atom changed the world but once again man decided to...

ipcamtalk.com

Regarding a camera for plates (LPR) - keep in mind that this is a camera dedicated to plates and not an overview camera also. It is as much an art as it is a science. If you want to capture plates, you need a camera just for that.

During daytime it may act as an overview, but at night the shutter tends to need to be faster to capture the plate, and the faster the shutter, the darker the image.

Most of us run B/W to take advantage of the reflective properties of a plate which is why we can run a faster shutter to get the plates. Once you decide to keep it in color, the shutter speed slows way down and might not be fast enough to actually capture a plate.

You may have enough light and the car moving slower to capture, but you may also need to decide what is more important - the overview or the plate. Most of us run two cameras.

At night, we have to run a very fast shutter speed (1/2,000) and in B/W with IR and the image will be black. All you will see are head/tail lights and the plate. Some people can get away with color if they have enough street lights, but most of us cannot. Here is a representative sample of plates I get at night of vehicles traveling about 45MPH at 175 feet from my 2MP camera (that is all that is needed for plates):

View attachment 111403

Appreciate the reply. I have been working on getting my new instance of BI installed and configured with DI and now that I have it up and running per (youtube.com/watch?v=nLH9GEcdb9Y) which was highly recommended as a starting point in the BI forum (blueirissoftware.com/forum/viewtopic.php?f=11&t=2426) I would like to touch on some of your prior comments.

I updated to 5.4.4 and it seems that the stored videos are much smaller in size prior to the new version which is good news. Prior to update was averaging 1gb or higher, now in the 25-150mb range.

I turned of face recognition as it doesnt seem to work (no new faces id'd) and cameras are not placed in the ideal location. I am not using LPR and probably wont unless needs change.

I am working with the new instance and have several questions that I will post once I run out of steam and hit the wall. Appreciate the support, stay tuned for more.

 

[looney2ns]

I highly recommend that you just stick with the learning curve that BI needs, before you dive into Deepstack.
Take the time to read the included BI help file, and look at their help videos.
Blue Iris Support - YouTube

 

[wittaj]

We have seen face recognition not work very well at all as it relates to DeepStack - there are a lot of quirks and hoops to jump through to try to make it work. Unless it got fixed recently, you cannot run object detection and facial at the same time for example. I am not aware of anyone on this forum that says it is working as intended.

Even if the DeepStack portion was working as intended, most of us have found that facial recognition is more gimmicky and novelty than anything else. If you have to put in 5 or 10 or 15 or 35 pictures of yourself in system for it to recognize it is you...then you shouldn't expect much.... My success rate was under 5% so I moved on to other things LOL.

It can work in certain situations like a business that requires everyone to stop in front of the camera and the camera is at head height. Outside of that, the percentage of being accurate is probably not going to be super high. Especially someone coming to a front door.

Unless you spend the big bucks that casinos and airports have LOL.

Heck even in ideal situations like a business with the camera at ideal height and optimal lighting it fails....

What do you think of Dahua's "Uyghur Warnings"?

Good luck. Wish you no ill will. I'm not here to argue with people simply to contribute when and where I can. This is also followed by lending my views on current technology and how it impacts all of us. It's undisputed that splitting of the atom changed the world but once again man decided to...

ipcamtalk.com

 

[rjp1267]

Anyone having issues with clips in web ui not being updated in a timely manner? I will see current clips in the admin view but there is a delay when those clips are visible in the web ui, sometimes by as much as a couple of hours.

Also if I wanted to configure the admin view to use HD stream rather than SD, cant stand the grainy look and slow response, does this place stress on the cameras or something else? FYI- I refer to the desktop version of the BI system as the admin view.

Merry Christmas to all.

 

[rjp1267]

Hello all...I have been getting acquainted with my setup and even found a use for 2 logitech usb webcams I had lying around. I have had an occasion or two when a camera was not responding in the sw but when i connected directly to the cam all things checked out, so i rebooted BI and all was as it should be. Still adding a camera or two and fine tuning detection but as I get more acquainted with and BI keeps improving on the software this is turning out to be a fine setup. FYI..Dell had a nice deal on a Vostro 5890 for <$600 with 512 ssd, 8gb ram which I added another 8gb with an i5 processor and it runs right in the sweet spot with <5% cpu and <40% mem usage, still enough wiggle room to add on another service. Unless I had seen a Dell system come up in my search I would have never even thought of them which is odd as I have 3 Dell monitors sitting right in front of me and I have always had good luck with Dell. I would highly recommend anyone looking for a new system to give them a thought, prices are higher now due to, well I wont get into politics, but they run sales frequently during the year so you can catch a deal. Thanks for all of your help and look forward to seeing you on another discussion thread.

 

[SandmanXX]

Curious, are you using a POE switch? if so, which one?