5 ms·

All leading camera brands have on SoC motion detection (Honeywell, Motorola...) and they have ML driven detection on their on premise stream servers. This is go

by avip 7y ago

All leading camera brands have on SoC motion detection (Honeywell, Motorola...) and they have ML driven detection on their on premise stream servers. This is going to be a very tough sell.

taway555 7y ago
I'm in the industry and I have to agree here. All of the devices we develop and use (third-party) either already have this capability built-in to the device or are working on it right now.
- mrafiee 7y ago
  I agree it is an active area and other people are working on the problem, but can you be more specific about the capabilities that you have seen in other products?
  - taway555 7y ago
    Essentially all of the events you mention plus a good amount more are either in an active product development pipeline for vendors we work with or are actively being worked on internally on devices we are building. It's a tough space.. margins on these hardware devices are thin already, and sometimes negative as they are simply loss leaders hooking in users for long-term saas payments, so in our case, it's cheaper to build it in-house. Good luck, however!
    - mrafiee 7y ago
      I completely agree with you re margin on the hardware being low. We are not planning on making any money on the hardware. But I also believe it is not possible for startups/companies that don't have copious resources (like Google and Amazon) to do great at both hardware and AI (i.e. build cutting edge AI/CV tech)... Thank you :) If you'd be up for discussing this further, may I ask you to email me? (rafiee at visualone.tech)
mrafiee 7y ago
There are various cameras on the market claiming to have smart alert features but in reality the only AI features that they are currently offering (that we are aware of) are person detection, unfamiliar face detection, facial recognition, activity zones, some limited object detection... if you know of any solutions offering beyond those, we would love to know about them... also these features come at a high price both in terms of the initial hardware cost and monthly fees (e.g. Nest Cam IQ selling for $300-$400)
- gbrown 7y ago
  One super annoying thing I've noticed about OpenCV - no ability to read relevant metadata from streams. It seems like lots of applications of OpenCV would benefit from using the motion capture functionality of these cameras (for example, streamed over RTSP via OnVIF) as a pre-screening technique, but I've had trouble getting access to it. It seems like it should be straightforward to add, but I haven't had a chance to do so.
  - mrafiee 7y ago
    what metadata specifically are you trying to get from the streams?
    - gbrown 7y ago
      The OnVIF specification allows data on motion to be transmitted with frames as metadata. This is really useful for, for example, IP cameras.
- mike_d 7y ago
  Axis has super advanced AI/ML video stuff. I've seen demos of it doing things like "use a network of PTZ cameras to auto track the person who left an unattended bag". They have lots of free applications in their App Gallery, and with minimal effort cheap Chinese cameras can be made to work with the system (since Axis OEMs their cameras from them anyway). https://www.axis.com/en-us/products/camera-applications/application-gallery https://www.axis.com/en-us/products/camera-applications/appl...

by avip 7y ago

All leading camera brands have on SoC motion detection (Honeywell, Motorola...) and they have ML driven detection on their on premise stream servers. This is going to be a very tough sell.

taway555 7y ago

I'm in the industry and I have to agree here. All of the devices we develop and use (third-party) either already have this capability built-in to the device or are working on it right now.

mrafiee 7y ago
I agree it is an active area and other people are working on the problem, but can you be more specific about the capabilities that you have seen in other products?
- taway555 7y ago
  Essentially all of the events you mention plus a good amount more are either in an active product development pipeline for vendors we work with or are actively being worked on internally on devices we are building. It's a tough space.. margins on these hardware devices are thin already, and sometimes negative as they are simply loss leaders hooking in users for long-term saas payments, so in our case, it's cheaper to build it in-house. Good luck, however!
  - mrafiee 7y ago
    I completely agree with you re margin on the hardware being low. We are not planning on making any money on the hardware. But I also believe it is not possible for startups/companies that don't have copious resources (like Google and Amazon) to do great at both hardware and AI (i.e. build cutting edge AI/CV tech)... Thank you :) If you'd be up for discussing this further, may I ask you to email me? (rafiee at visualone.tech)

mrafiee 7y ago

There are various cameras on the market claiming to have smart alert features but in reality the only AI features that they are currently offering (that we are aware of) are person detection, unfamiliar face detection, facial recognition, activity zones, some limited object detection... if you know of any solutions offering beyond those, we would love to know about them... also these features come at a high price both in terms of the initial hardware cost and monthly fees (e.g. Nest Cam IQ selling for $300-$400)

gbrown 7y ago
One super annoying thing I've noticed about OpenCV - no ability to read relevant metadata from streams. It seems like lots of applications of OpenCV would benefit from using the motion capture functionality of these cameras (for example, streamed over RTSP via OnVIF) as a pre-screening technique, but I've had trouble getting access to it. It seems like it should be straightforward to add, but I haven't had a chance to do so.
- mrafiee 7y ago
  what metadata specifically are you trying to get from the streams?
  - gbrown 7y ago
    The OnVIF specification allows data on motion to be transmitted with frames as metadata. This is really useful for, for example, IP cameras.
mike_d 7y ago
Axis has super advanced AI/ML video stuff. I've seen demos of it doing things like "use a network of PTZ cameras to auto track the person who left an unattended bag". They have lots of free applications in their App Gallery, and with minimal effort cheap Chinese cameras can be made to work with the system (since Axis OEMs their cameras from them anyway). https://www.axis.com/en-us/products/camera-applications/application-gallery https://www.axis.com/en-us/products/camera-applications/appl...