Hop on the serverless adventure

on YouTube ↗44m

" summary

Shows a working pipeline that scrapes images with the Bing Image Search API, pushes URLs onto a queue, then trains a custom model via Azure Custom Vision using an Azure Functions serverless job. Requires at least five images per tag, one demo project has ~169 images, model reports per-tag precision (example: 100% for one tag, 33% for a random tag) and a live prediction returned 96% for “Marty McFly.” Notes limits: function timeouts (minutes) and cold starts affect latency.

AI-written summary · about

" chapters (16)

1. 0:00Start
2. 7:11UNSUPERVISED
3. 11:41GOAL
4. 15:09CUSTOM VISION
5. 15:49BUILD A CLASSIFIER
6. 21:33EXPORT MODEL TO MOBILE
7. 26:07TURBO
8. 26:32Execution time limit
9. 27:01Functions are stateless
10. 27:15Cold start
11. 28:37Serverless providers
12. 28:57Hello world
13. 29:31Timer
14. 29:45Web hook
15. 31:27Data processing
16. 41:41Resources

" related talks

• 23mBuild a look alike engine with machine learning and AngularSimona Cotin · NGConf 2018
• 55mAzure Cognitive Search Beyond the First DemoMarco Minerva · Global Azure Virtual 2020
• 58mPlaying the game with Azure Cognitive Servicespeter von eden · Global Azure Virtual 2020
• 01h 04mBuilding and Training your own Custom Image Recognition AIDavid Giard · NDC Oslo 2018
• 01h 00m5 Levels of AI in Microsoft AzureAlan Smith · CloudBurst 2022
• 26mLet's Build a Doppelganger Game!Simona Cotin · NodeConf EU 2018
• 01h 01mGetting Started with Computer VisionSeth Juarez · NDC London 2021
• 16mHow to get started with Computer Vision AutoML in AzureGavita Regunath · SQLBits 2023