AI in the battle against fakes
summary
A smartphone-photo system is built to decide genuine versus counterfeit packaging by combining per-icon object detection and dedicated logo classifiers rather than one big model. It uses Azure Functions (Durable) to orchestrate, Umbraco for product config pushed to Cosmos DB and Blob Storage, Docker APIs hosted from Azure Container Registry and CI/CD with Azure DevOps. Custom Vision does object detection, Azure Machine Learning trains a bespoke Philips-logo classifier, and practical thresholds sit around 80–90% confidence for acceptance. The modular pipeline is deployable and operational.
AI-written summary · about
chapters (17)
- 0:00Start
- 5:38Ingredients
- 7:01Umbraco config
- 12:09Icon validator API
- 13:35Icon detection
- 15:01Icon verification
- 18:01Object Detection
- 18:13Custom Vision Service
- 29:36Logo classification
- 34:23Setup your environment
- 35:14Azure Machine Learning Service
- 37:14Create a workspace
- 39:46Create an experiment
- 40:02Create a training file
- 40:25Create an estimator
- 50:14Register the model
- 56:30Recap
related talks
48mAI in the battle against fakes
59mAI in the Battle Against Fakes
42mProduct Recognition in Retail/E-Commerce with Azure AI
42mProduct Recognition in Retail/E-Commerce with Azure AI
01h 04mBuilding and Training your own Custom Image Recognition AI
41mAI + Machine Learning for .NET Developers
36mAI-Powered Real-Time Quality Control
50mAzure AI, Power new possibilities for every organization