MV Sense settings enabled, Tensoflow model installed on camera, webhook in place but still no real t

Solved
AmeyGadre
Here to help

MV Sense settings enabled, Tensoflow model installed on camera, webhook in place but still no real t

Hello All,

 

Recently I enabled MV Sense settings for a specific camera, installed a valid tensorflow model accepted on meraki dashboard.

Set up webhook to get events, did the test webhook on which I can see logs in my webhook API.

 

But when it comes to actual detection events ( object detection ) it does not trigger any event to webhook.

 

Is there a setting I am missing? Any help would be highly appreciated.

 

Thank you 

1 Accepted Solution
alemabrahao
Kind of a big deal
Kind of a big deal

1 - You can configure inference zones directly from the Meraki Dashboard, if you have the necessary camera permissions. Go to the specific camera > Settings > Sense > Inference Zones

 

2 - Yes, it is.

I am not a Cisco Meraki employee. My suggestions are based on documentation of Meraki best practices and day-to-day experience.

Please, if this post was useful, leave your kudos and mark it as solved.

View solution in original post

9 Replies 9
alemabrahao
Kind of a big deal
Kind of a big deal

What is your camera model?

I am not a Cisco Meraki employee. My suggestions are based on documentation of Meraki best practices and day-to-day experience.

Please, if this post was useful, leave your kudos and mark it as solved.
AmeyGadre
Here to help

Its MV93M

alemabrahao
Kind of a big deal
Kind of a big deal

The MV93 needs one or more inference zones configured so it knows where to detect. No zones = no events. (Zone-less models do not auto-detect across the full frame.)

Make sure your payload template includes alertData, and your webhook filters allow the right event types.

 

https://www.youtube.com/watch?v=1i80QMq1X-g

I am not a Cisco Meraki employee. My suggestions are based on documentation of Meraki best practices and day-to-day experience.

Please, if this post was useful, leave your kudos and mark it as solved.
AmeyGadre
Here to help

Thank you!! two questions:

1. This inference zones configuration is something I believe network team needs to do who installs the cameras or is that something we can handle via meraki dashboard?

2. Currently I am using default meraki payload template, so I believe that should be good enough. 

alemabrahao
Kind of a big deal
Kind of a big deal

1 - You can configure inference zones directly from the Meraki Dashboard, if you have the necessary camera permissions. Go to the specific camera > Settings > Sense > Inference Zones

 

2 - Yes, it is.

I am not a Cisco Meraki employee. My suggestions are based on documentation of Meraki best practices and day-to-day experience.

Please, if this post was useful, leave your kudos and mark it as solved.
AmeyGadre
Here to help

Thank you Alessandro!! appreciate your help in this matter.

AmeyGadre
Here to help

.

AmeyGadre
Here to help

I enabled and performed all of the above steps, here is my observation:

 

1. When I enable sense api & upload custom cv - tensorflow model - Motion alerts tab gets disabled.

2. For some reason adding inferences zones - area and line - it still did not trigger any event.

3. As soon as I disable Custom CV - I can see Motion Alerts tab gets enabled - and I can also get motion alerts when object crosses inference zones.

 

My question here would be: Does this mean there is some issue with tf model that I have on that camera? or something might still be missing from configuration perspective. Thank you

oliviaaaaa
Just browsing

Have you checked that the MV Sense camera's object detection is enabled for the correct event types and that the webhook is configured in the (Notifications and Events) section?
Test webhooks do not always reflect real-time detection if event triggers are not fully active.

Get notified when there are additional replies to this discussion.