Eldercam: Event-driven AI for Surveillance
summary
Builds a live system that captures one photo per minute from an esp32 camera, uploads to MongoDB Atlas, triggers events to AWS EventBridge and a Python Lambda that calls AWS Rekognition, then writes labeled metadata back and notifies via Twilio or SMS on exceptions. Uses Atlas triggers, serverless functions, time‑series window queries for 15‑minute maxima, charts and a 48‑hour retention policy (48 hours). It works end‑to‑end; a content‑type bug was found and fixed.
AI-written summary · about
chapters (14)
- 0:00Intro
- 3:00The rough setup
- 6:38MongoDB document databases
- 8:51The MongoDB query API
- 12:30Arduino IDE
- 13:58Realm browser SDK
- 15:14Atlas serverless function
- 23:49JSON has limits
- 25:10EJSON
- 26:07HTTPS endpoints
- 28:22Alerts on events
- 30:48Daily roundup
- 35:41Tidy up
- 37:58Outro
related talks
37mUnstructured Data Processing with a Raspberry Pi AI Kit and Python
29mRekognition Live Demo
32mThe ML system endeavors
26mServerless Architecture for Product Defect Detection
13mSpotting and solving everyday problems with machine learning
45mBuilding Interactive, AI-powered IoT Applications
01h 06mAmazon Machine Learning to Predict Responses
24mCreating a Smart and Serverless Cat Detection System with IoT and Image Recognition APIs