Overview
Binsense audits waste streams using vision models to catch recycling contamination before trash reaches the landfill. Commercial buildings and college campuses use it to identify mismanaged bins, track sorting habits, and adjust collection routes based on real container fill levels.
Technical implementation
- Vision classification. Gemini vision models review photos taken at disposal stations. The system categorizes items into plastics, cardboard, organics, or hazardous materials, assigning confidence scores and flagging foreign objects in recycling bins.
- Analytics dashboard. A Next.js interface maps contamination levels floor by floor and bin by bin, producing weekly compliance summaries and pinpointing where recycling rules are routinely ignored.
- Storage and alerts. Supabase handles Postgres data persistence with row-level security to isolate tenant properties. Real-time websocket subscriptions push immediate alerts to staff when a container exceeds contamination limits.
Practical impact
Manual waste audits normally involve hours of clipboard logging and hand sorting. Binsense produces clean audit numbers within seconds of an image upload, helping operators cut hauling fines and place signs where people actually need them.