EcoScan – Sustainable Fashion Scanner
Timeframe: Nov 2025
Award: 🏆 1st Place — Sustainability Track, HackNYU 2025 (Major League Hacking)
Stack: React Native · Expo · TypeScript · FastAPI · Supabase (Postgres) · Google Cloud Vision · Google Gemini 2.5-flash · Lykdat API
Overview
Built a mobile app that scans clothing tags and items, returning an Eco-Score (0–100, A–F grading) with AI explanations and up to 40 greener alternatives per product. Won 1st place in the HackNYU 2025 Sustainability Track.
Scan Pipeline
The pipeline runs in 15–30 seconds end to end:
- Lykdat API deep-tags the clothing image for item type, colors, patterns, brand (fallback: Vision OCR + Gemini)
- Cloud Vision reads label text for material composition, country of origin, certifications
- Heuristic weighted eco-score formula combines material impact matrix, origin labor/transport scores, and certification bonuses (GOTS, Fair Trade, B Corp)
- Lykdat Global Search finds ~40 visually similar retailer products
- Each alternative is deep-tagged, eco-scored, and top 5 returned
Key Features
- Two-stage camera scan: separate captures for the clothing and its tag
- Impact breakdown: per-item scores for microplastics, carbon, water, and labor risk
- Personalized picks: Gemini analyzes scan history to recommend products aligned with user style and sustainability preferences
- Scan history and trends tracked via local AsyncStorage and Supabase