What is boundingBoxGrid in analyticsLookupObjectTracks?

Answered

Follow

akshay thoughtclan

• April 06, 2020 05:39

I am working on object detection analytics part in network optix. I came across this API to fetch object analytics data. somehow, I could not understand  boundingBoxGrid. What data it has? How to decode this?

My data has like below,

{
"attributes": [
   {
   "name": "Sex", 
   "value": "Female"
   },
   {
   "name": "Name",
   "value": "Judy"
   },
   {
   "name": "Hair color",
   "value": "Blonde"
   },
   {
   "name": "Age",
   "value": "42"
   }
   ],
   "bestShot": {
      "rect": {
          "height": 0.3988280534700604,
          "width": 0.30366843679423794,
          "x": 0.5716901666361472,
          "y": 0.3504242202282854
           },
       "timestampUs": "1586148469184000"
       },
   "deviceId": "{e3e9a385-7fe0-3ba5-5482-a86cde7faf48}",
   "firstAppearanceTimeUs": "1586148468184000",
   "id": "{ef3cc442-39f2-43d1-a61a-fbb0e5cc1770}",
   "lastAppearanceTimeUs": "1586148470151000",
   "objectPosition": {
       "boundingBoxGrid":                        "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAf/wAAH/8AAB//AAA//wAAP/8AAD//AAA//wAAP/8AAD//AAA//wAAP/8AAD//AAA//wAAP/8AAD//AAA//wAAP/8AAD//AAA//wAAAAAAAAAAAAAAAAA="
},
  "objectPositionSequence": [],
 "objectTypeId": "nx.stub.humanFace"
}



I am interested in this part,

"objectPosition": {
"boundingBoxGrid": "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAf/wAAH/8AAB//AAA//wAAP/8AAD//AAA//wAAP/8AAD//AAA//wAAP/8AAD//AAA//wAAP/8AAD//AAA//wAAP/8AAD//AAA//wAAAAAAAAAAAAAAAAA="
}

2

• 

  Sergey Yuldashev

  • April 08, 2020 11:16
  • Edited

  Hello Akshay,

  The "objectPosition" attribute shows the "trace" of the object across the whole camera field of view across the whole time while object was being detected.
  What you see here is basically a bit mask "encoded" as a QbyteArray which is mostly designed to be used by Nx Client.

  If you need to get the location of the object I would recommend you to use the "bestShot" attribute.
  Basically it shows you the coordinates of the object in space and time in the human readable format at the moment when the object was most distinguishable and sharp. Which is the best option for actual operators.

  0

  Comment actions Permalink
• 

  Mohamed Thasin

  • April 08, 2020 11:18

  thanks for the response. So, Can I ignore the objectPosition attribute? 

  0

  Comment actions Permalink
• 

  Sergey Yuldashev

  • April 08, 2020 11:57

  Yes, it is ok to ignore that.

  0

  Comment actions Permalink

Please sign in to leave a comment.