Impacting Global Policy by Understanding Litter Data
summary
Data-backed claims, not slogans: a dataset of about 15 million images from 185 countries feeds computer vision and ML that labels object, material and brand, estimates weight/volume and even decomposition age. The pipeline runs on serverless step functions and handles spikes (≈30 uploads/sec on World Cleanup Day). Models power predictive litter-density maps, CCTV and drone ingestion, and have driven policy changes—San Francisco’s cigarette tax (≈$4M/yr) and municipal bag bans—showing measurable before-and-after effects.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:14What is Litter
- 5:19The Scale of Litter
- 7:38Consequences of Litter
- 10:45What do we see
- 12:38Our data
- 18:38Location specific models
- 22:45Whats next
- 24:48Alternative sources
- 26:47Questions
- 29:21Types of Litter
related talks
28mData Curation: Transforming Bytes into AI Gold
13mApplying Computer Vision to Reduce Contamination in the Recycling Stream
26mThe Data I Breathe
36mWatching Millions of Trees with Python: For Impact and Profit
22mUrban Artificial Intelligence in Amsterdam
21mBuilding sustainable communities through interaction
09mSocial Media Image Analysis for Public Health
43mPython at City Scale