On-premise version

N

netmax

Getting the hang of it
Sep 6, 2015
30
25
Hi,

I'm using BlueIris since years and I love this product, pretty high on my wishlist is some people/face/car recognition feature and I was excited reading about the Sentry thing in the first five minutes ... until I've read that it's a cloud thingy :(

I don't really like cloud services where any of my private data is sent to. I'm nearly running everything at home on my servers (Seafile, Alfresco, Jira, IceWarp) and my home CCTV has pretty much the highest privacy rating, even all my cams are blocked in my firewall to deny any connection to the outside world. Call me strange, but that's the way I like it and it has proven to work extremely well over the last years.

I'm not gonna send *any* frames to some cloud servers for AI, whether they are encrypted or not - in the end they are ending up somewhere on a machine as the need to be processed in clear.

I'd really love to have that feature, the price tag of $50 per year is fine - but on my own servers, please.

I've read somewhere here that an on-premise version is on the roadmap and I'd love to see that rather sooner than later. Preferred running on Linux ... not another crappy Windows VM on my servers.

@varghesa: If this is going to beta please feel free to PM me if you plan to do a public beta.
 
  • Like
Reactions: social26 and keneil01
Great feedback. There are and will always be customers that are only interested in on-prem solutions. We respect that and hope to provide a solution with BI. FYI, all our inference models run on Ubuntu. :)

I will keep you in mind for future beta testing when an on-prem solution is closer to a market solution. Unfortunately, right now it is a more later than sooner release.

-- Sam from Sentry
 
You are paying $50 per year, Skinny1, if you can make something better, why don't you? Or maybe troll is strong with you today
 
I ran Sentry thru the trial period but ultimately decided against it.

There were a few reasons:
1) This is the big one.....I don't want everything on my cameras going up to "The Cloud" to be analyzed. I know that's tin foil hat paranoid and I don't have any indoor cameras but I just decided that I don't want them to leave my house.
2) While $50/yr isn't a bad price I'd have to add another $20/yr to that for the extra 3 cameras that I have (13 total) and I just don't see it as being worth $70/yr ****to me****.
3) I was getting quite a few false positives (dogs mainly as well as a bunch during our recent storms). It was much better than plain motion detection though but it would also miss people sometimes as well. It's still a very impressive product though.
4) I do actually want to see when vehicles pull into my driveway but not really birds walking around or flies that land on the camera.

I plan on re-evaluating in the future.
 
  • Like
Reactions: Arjun, t_andersen, hdcam01 and 1 other person
The other side of the coin.

I also ran Sentry thru the trial period but ultimately decided for it.

There were a few reasons:
1) I'm not tin foil hat paranoid and I have any indoor cameras that are disabled when home.
2) $50/yr isn't a bad price, I have (12 total) and I just left 2 off the Sentry system.
3) I was getting a few false positives and is much better than plain motion detection. It's a very impressive product.
4) I will still see when vehicles pull into my driveway (one of the 2 cameras not using Sentry.

I also plan on re-evaluating in the future.
 
  • Like
Reactions: SentryAI
Freedom to choose is a wonderful thing :headbang:

If I could have the detection done locally and be able to select vehicles as well as people I'd be more than happy to pay the $70/yr.

For now though I'll just live with standard motion detection alerts from the 4 main perimeter cameras. The bird and fly alerts to get old though while the neighbor dog is at least entertaining to watch while she runs around my yard throughout the day :lol:
 
  • Like
Reactions: social26, SentryAI and anijet
I have 16 cameras going through BlueIris & Sentry and have been using Sentry extensively since it was released (2 months ago?).

1) Detection of humans is probably close to 98% accurate. I've had my large German shepherd detected as a human maybe 10 times out of the 1,000+ alerts he created.
2) I disable Sentry for my indoor cameras so I'm not concerned about privacy.
3) If a Sentry programmer got off of me walking up to my house, well I feel sorry for them since I'm ugly lol. Seriously though, the outdoor cameras won't capture anything highly "sensitive" other than maybe me scratching my butt as Im walking to the door. Point is, yeah its private data but Im not sending any really private stuff to them.
4) $50 per year is too cheap imo. For the features we are requesting and getting, they're not really charging enough. Maybe if we paid more, they could hire more devs and push out the features we want faster, maybe not. Im in favor of paying for a good service, which I feel we are getting.
5) My daily false alerts went from probably 300-500 down to virtually zero. My house is surrounded by 50 palm trees that make massive moving shadows, no amount of Blue Iris motion tweaking, zone crossings...etc will come remotely close to what Sentry does.

Anyways it seems like Sentry is focusing on the most requested features from the public, which I happily applaud them!
 
  • Like
Reactions: Arjun, social26, kwijibo and 2 others
BlueWave said:
I have 16 cameras going through BlueIris & Sentry and have been using Sentry extensively since it was released (2 months ago?).

1) Detection of humans is probably close to 98% accurate. I've had my large German shepherd detected as a human maybe 10 times out of the 1,000+ alerts he created.
2) I disable Sentry for my indoor cameras so I'm not concerned about privacy.
3) If a Sentry programmer got off of me walking up to my house, well I feel sorry for them since I'm ugly lol. Seriously though, the outdoor cameras won't capture anything highly "sensitive" other than maybe me scratching my butt as Im walking to the door. Point is, yeah its private data but Im not sending any really private stuff to them.
4) $50 per year is too cheap imo. For the features we are requesting and getting, they're not really charging enough. Maybe if we paid more, they could hire more devs and push out the features we want faster, maybe not. Im in favor of paying for a good service, which I feel we are getting.
5) My daily false alerts went from probably 300-500 down to virtually zero. My house is surrounded by 50 palm trees that make massive moving shadows, no amount of Blue Iris motion tweaking, zone crossings...etc will come remotely close to what Sentry does.

Anyways it seems like Sentry is focusing on the most requested features from the public, which I happily applaud them!
Click to expand...


Very kind words. Thank you.

-- Sam from Sentry
 
varghesesa said:
Very kind words. Thank you.

-- Sam from Sentry
Click to expand...

I tried a few days when it first released but it didn't generate any alert, but not sure if I set it up correctly.

One question I always wondered, should I remove all my zone alert and let BI motion triggerred as frequently as possible but let Sentry handle the alert decision? Wish I can have another chance to try it with proper settings.
 
social26 said:
I tried a few days when it first released but it didn't generate any alert, but not sure if I set it up correctly.

One question I always wondered, should I remove all my zone alert and let BI motion triggerred as frequently as possible but let Sentry handle the alert decision? Wish I can have another chance to try it with proper settings.
Click to expand...

I was at first confused by the right way to move to forward. I believe the consensus now, which I agree with it as well is to keep all the zone alerts & activate Sentry AI for human detection.

Think about the trigger dialogue for Motion / zone+hotspots / Sentry as a trigger pipeline (motion --> zone/hotspots --> Sentry). First define the motion. If there is no motion, then there is no reason to trigger. Then define your zones + hotspots which are basically regions of interest for said motion. So if motion is outside of area of interest, the motion is suppressed. Finally, let Sentry AI determine human activity.

Some users were using zones to reduce false alerts. In those cases, you could probably simplify the logic. For example, to reduce false alerts, users would cut their driveway into areas and trigger only if motion crossed areas. This reduced false alerts due to tree shadows on the driveway. In this case, you could probably reduce the zones to just one for the driveway and let Sentry do all the work to determine human activity on the driveway.

In the next integration, we plan to more tightly couple Sentry AI to the areas of interest defined by BI. Currently, we examine the entire image. For example, when we release vehicle detection, users probably do not want to be alerted to parked cars on the street or cars driving along the road. Thus, users would create a driveway zone and alerts would only be sent if vehicle motion was detected on the driveway.

-- Sam from Sentry
 
  • Like
Reactions: jaydeel, social26, petexa and 3 others
varghesesa said:
I was at first confused by the right way to move to forward. I believe the consensus now, which I agree with it as well is to keep all the zone alerts & activate Sentry AI for human detection.

Think about the trigger dialogue for Motion / zone+hotspots / Sentry as a trigger pipeline (motion --> zone/hotspots --> Sentry). First define the motion. If there is no motion, then there is no reason to trigger. Then define your zones + hotspots which are basically regions of interest for said motion. So if motion is outside of area of interest, the motion is suppressed. Finally, let Sentry AI determine human activity.

Some users were using zones to reduce false alerts. In those cases, you could probably simplify the logic. For example, to reduce false alerts, users would cut their driveway into areas and trigger only if motion crossed areas. This reduced false alerts due to tree shadows on the driveway. In this case, you could probably reduce the zones to just one for the driveway and let Sentry do all the work to determine human activity on the driveway.

In the next integration, we plan to more tightly couple Sentry AI to the areas of interest defined by BI. Currently, we examine the entire image. For example, when we release vehicle detection, users probably do not want to be alerted to parked cars on the street or cars driving along the road. Thus, users would create a driveway zone and alerts would only be sent if vehicle motion was detected on the driveway.

-- Sam from Sentry
Click to expand...
This is very pertinent information that should be given when you initially activate Sentry. I searched for a while to find this info. Your "checklist" is extremely basic and I was wondering if should reconfigure my zones and crossing entries to allow Sentry to work more efficiently. Thanks
 
  • Like
Reactions: varghesesa
If the community would like I can make this for you, the only requirement is Python needs to be installed on your windows machine... It's quite easy to do and I've had good results with it detecting people and cars, dogs, cats, boats, etc, to make relevant web requests. It uses the Yolo V3 object detection model which is about 250mb in size. You can specify the "confidence" threshold as well.

It's open source and has no license restrictions.

I'm just not sure what people want to achieve generally... I can't make it properly Integrate with BI, but it can check alert images and make a web request (even to BI's web server is possible). It gives responses in half a second, and only runs when triggered.

What would be the ideal functionality for this?

Keep in mind, if a vehicle is parked in view of a camera, any alert trigger caused by something else will pass a vehicle detected response.
 
varghesesa said:
Sorry, on-premise is not a 2019 roadmap item. A customer portal so users can see, delete and report false negatives/positives on their images is part of the 2019 roadmap.
Click to expand...

A portal of all our pictures is not necessary when BI has it's own storage for this. Why not let us do this on our end local vs. building a portal we have to log into.
 
  • Like
Reactions: varghesesa
integlikewhoa said:
A portal of all our pictures is not necessary when BI has it's own storage for this. Why not let us do this on our end local vs. building a portal we have to log into.
Click to expand...

Customers have been requesting ways to see which image is sent to Sentry so they can tweak parameters to improve performance and ways to delete and report false negatives/positives from within BI. All good ideas for future versions.

-- Sam from Sentry
 
  • Like
Reactions: anijet, pmcross and fenderman
I"m very much interested in this as well. Any progress on this.
I read camect is able to do AI locally on their celeron box. Something similar for BI?
 
You must log in or register to reply here.
Top Bottom