Anomaly detection in the cloud: how to detect issues in data [...]
summary
Practical walkthrough of using AWS managed APIs Amazon Lookout for Vision and Lookout for Metrics to detect image defects and time‑series anomalies without ML expertise, training an image model in about one to two hours from 30–50 images (demo used ~1,000) and often reaching 100% accuracy on simple industrial defects; metrics detector connects to S3/MySQL/GA, produces a single severity score across dimensions and surfaces correlated root causes (country, platform), and vision models incur hourly billing so start/stop to control cost.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 3:00What is anomaly detection
- 4:30What is anomaly detection in practice
- 5:10What kind of anomalies do we meet
- 8:35What are APIs
- 10:10What are services
- 11:30Vision
- 12:15Data set
- 13:30What is anomaly
- 18:40Examples
- 20:00The data set
- 21:40Creating a detector
- 23:50Mapping
- 25:03Pretrained model
- 30:25Outro
related talks
29mAmazon Lookout for Vision
26mServerless Architecture for Product Defect Detection
28mDetect abnormal equipment behavior by analyzing sensor data
28mProactively monitor health of your business using Amazon Lookout for Metrics
51mAnomaly Detection as a Service - What is normal?
29mAI-Powered Anomaly Detection in Cloud Data Pipelines
27mComputer vision for automated quality inspection
30mBetter data quality with ML & visualization