Revive
Machine Learning powered recycling assistant that recognizes items from camera photos and maps nearest certified drop-offs.
Timeline
2019
Role
Full Stack Mobile Development
Context
CUNY Hackathon 2019
Platform
React Native (iOS & Android)
Google Cloud Platform Vision APIInspiration
“New York City has no landfills or incinerators within its borders, yet residents produce over 12,000 tons of waste every day. What truly happens when you throw something away?”
We wanted people to have an immediate, actionable answer: give it to ReViVe. Built during an intense 48-hour hackathon, Revive tackles the confusion of municipal waste sorting by removing guesswork. Instead of wondering whether a plastic container or broken appliance can be recycled, users simply take a photo.
Capture Photo
Live camera capture analyzed via Google Vision API
Verify Category
ML classification output reviewed on device
Navigate & Drop Off
Closest verified facility mapped with intake hours
Core Features
Visual Item Classification
Instant photo recognition powered by Google Cloud Vision API, identifying object materials (plastics #1–7, aluminum, cardboard, e-waste).
Interactive Confirmation Sheet
A user review screen allowing instant adjustment of detected categories before querying municipal and specialized recycling facilities.
Recycling Inventory List
Capture multiple items across a household clean-out; items are saved locally so users can drop them all off in a single optimized trip.
Proximity Center Mapping
Automated radius lookups combining Earth911 and NYC Open Data registries to pinpoint nearest centers accepting the exact items photographed.
Facility Hours & Rules
Surface detailed center policies, phone contacts, accepted grades, and drop-off fees directly within the mobile view.
Offline Image Caching
Local image buffer management built on React Native and Expo, enabling smooth review even with intermittent subway connectivity.
Technical Capabilities & Architecture
| Capability | Implementation & Architecture |
|---|---|
| Mobile Framework | React Native |
| Computer Vision Engine | Google Cloud Vision API (Label & Object Localization) |
| Machine Learning Subsystem | TensorFlow.js feature extraction and confidence scoring |
| Cloud Infrastructure | Google Cloud Platform (GCP Cloud Functions & Storage) |
| Municipal Datasets | NYC Open Data & Earth911 Environmental Services API |
| Local Persistence | AsyncStorage & native image filesystem caching |
| Target Platforms | Cross-platform iOS and Android |
Hackathon Engineering & Learnings
- Overcoming Camera Subsystem Crashes: Resolved device-specific React Native camera permission and capture freeze issues across differing Android and iOS hardware configurations under tight 48-hour constraints.
- Deeply Nested Payload Ingestion: Constructed recursive data flattener routines to parse non-standardized multi-tier JSON payloads returned by legacy municipal environmental APIs.
- Cross-Campus Collaboration: Assembled a multidisciplinary team representing diverse CUNY campuses, pairing rapid UI prototyping in Expo with cloud image inference pipelines on GCP.
Built With
View Hackathon Submission
Read the original hackathon entry on Devpost or browse my full portfolio of products and experiments.