IoT and Facial Recognition at Scale: Using Amazon’s DeepLens to search for matches from the US’s Missing Persons Database
Andrew Carlson, Vicki Foss, Gerard Kelly, Michelle Liu
MIDS W251: Scaling Up! Really Big Data
University of California, Berkeley
August 2018
Repo for W251 final project, for which we have created a cloud-based system which uses the AWS DeepLens as an edge device to identify test subjects and pictures of people in the US's National Missing and Unidentified Persons System (NamUs) database, and which sends a text message alert with the person’s information upon finding a match.
-
IoT and Facial Recognition at Scale - Carlson, Foss, Kelly, Liu.pdf - Final paper
-
Project Presentation Slides.pdf - Slides used for in-class presentation of final project + results
-
src/ingest/ - Directory containing scripts for ingesting and indexing NamUs data
-
src/deeplens_face_detection.py - Lambda script for managing the DeepLens face detection
-
src/lambda_rekognition.py - Lambda script which checks for a match in the Amazon Rekognition index when the DeepLens detects a face and manages the alert process
-
rekog-demo.txt - Amazon Rekognition demo
-
stats - Metadata on the missing person files initially scraped from NamUs