NodeJS, ML, K8s and Unethical Face Recognition
summary
Demo links a NodeJS backend to Twitter profile images (followers filtered by ID) and runs face detection with face-api.js, loading ssd_mobilenet_v1 for accuracy and a tiny face detector option for speed. Images feed microservices in containers, queued through RabbitMQ, with sentiment scored via the sentiment npm package (±5) and results aggregated to a numeric “score.” Recognition works from a single reference image but fails on masks, avatars, or poor data and produces biased or mistaken scores (example: a demoed 16/100).
AI-written summary · about
chapters (14)
- 0:00Introduction and motivation
- 2:43Speaker background
- 4:29Data collection strategy
- 7:06Working with Twitter API
- 11:43Defining scoring criteria
- 13:27Machine learning approach
- 20:26Face detection techniques
- 23:22Performing face recognition
- 28:52Sentiment analysis
- 32:00System architecture
- 35:58Messaging queues
- 39:25Deployment with Kubernetes
- 43:47Live application demo
- 45:47Ethical considerations and wrap up
related talks
39mNodeJS, ML, K8s and Unethical Face Recognition
38mNodeJS, ML, K8s and Unethical Face Recognition
22mVue + TensorFlow.js: Building Your First ML Powered App
01h 05mIntro to machine learning in Node
35mAI JavaScript Rocks
54mIntroduction to machine learning in Node.js
48mThe Future of Machine Learning & JavaScript
39mAI + JavaScript Rocks