Skip to content

Latest commit

 

History

History
51 lines (43 loc) · 1.58 KB

Azure_functions.md

File metadata and controls

51 lines (43 loc) · 1.58 KB

Microsoft Azure

Custom Vision

Create

  • Name : Apple Recognizer

  • Subscription: Azure for students starter

  • Location: (Europe) West Europe

  • click on link and select the try for free option(Prediction pricing tier)

  • Prediction price choose F0 2 transactions per second

  • start a new project

  • Resource Group(AppleRECOG)

  • Training pricing choose F0

  • From the bell icon, Goto Resource and run specific service

  • Goto Custom Vison Portal from 2 point for basic theory

Working on another portal , seperately for Custom Vision

  • Further to work open customvision.ai
  • Create a new project
  • Give a name: AppleVSOrange
  • Description: Tells whether image is an apple or orange
  • resource: ApppleRecog[F0]
  • Project Types: Classification
  • Classification Types: MultiClass (used for single level classification)
  • Multilabel: not used here but generally for multi level classification
  • Domains: Food
  • Note : 5 images are required minimum per tag

Create a TAG

  • Just click on + icon on Tags
  • give a name
  • upload images of orange and add tag given-name
  • upload images

Train

  • click on green colored train icon on top of page
  • you find 2 ways ( Choose Training Type)
    1. Fast Training
    2. Advanced Training
  • Choose Advanced Training and time 3 hours

model info

  • Precision ~> tells you the rate of correct identification
  • Recall ~> correct identified / total identified
  • Go to Prediction tab
  • Start qucik test
  • upload image and check its probability

Azure Fundamentals