0

My Azure custom vision video object has high detection latency in the FO Tier. How can I minimize response time? Should I go for the S tier?

My plan used a custom vision object detection model which I trained on Azure custom vision portal to then use the prediction API in my Python script which sends a video frame by frame to an API. This has a lot of latency in response time. If I send a 1-minute video of 20FPS it takes 2+ hours to process it.

cthrash
  • 2,938
  • 2
  • 11
  • 10
Abdul Basit
  • 1
  • 1
  • Please explain a bit more about your problem. – Richie Bendall Mar 10 '19 at 09:57
  • thanks Richie Bendall for considering.I was using custom vision object detection model which i trained on Azure custom vision portal then use the prediction api in my python script which send video frame one by one to api but it take latency in rsponce time .If I send video of 1 min (20Fps) it take about 2or more hours in – Abdul Basit Mar 10 '19 at 10:05

1 Answers1

0

https://azure.microsoft.com/en-us/pricing/details/cognitive-services/custom-vision-service/ answers your question, but in general regardless what tier you are, you cannot send every frame to the service without extracting key frames first.

enter image description here

Ping Jin
  • 520
  • 4
  • 8