Keynote: Open Data @ Facebook
summary
Facebook states its open source program covers 500 projects with 300,000 contributions from 16,000 contributors and describes a year‑old Disaster Maps pipeline that produces five map types — population, movement, long‑term displacement, network coverage and inferred power — with strong privacy guards (anonymization, noise, smoothing, small‑count removal). A global ML model extracts roads from Maxar imagery, feeds a forked Rapid editor and an ML tasking manager (used in Paradise/Camp Fire showing evacuation to Chico), and the tools are open sourced.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 0:26Open Source Program
- 0:41Ecosystem
- 2:09Welcome
- 2:51Wildfires
- 4:00Critical Questions
- 4:56Partners
- 7:03Population Map
- 7:51Movement Map
- 8:23LongTerm Displacement Map
- 8:46Network Coverage Map
- 9:20Power Availability Map
- 10:20Maps at Facebook
- 13:08AI and Machine Learning
- 13:49Satellite Imagery
- 14:14Terrain Types
- 14:29Rural Areas
- 15:01ID Editor
- 15:51Rapid
- 16:07Humanitarian OpenStreetMap
- 16:59MaxR
- 17:27Walkthrough
- 18:45Local Communities
related talks
21mKeynote: HERE and the Open Location Journey
45mThe OpenStreetMap Project = Data + Community
26mOpen Source and the Volunteer Workforce
27mWhere Are My Keys? Unlocking Together What We Can't Alone
28mDelicious Democracy: snagging and mapping citizen data
12mOpen Data and the Future of ...
16mKeynote: Open Source as Endurance Infrastructure for Environmental Justice
12mThe O'Reilly Radar