Inspiration
E-waste is growing faster than our ability to manage it, yet most people still don’t know how to dispose of their old devices responsibly. I built EcoVision as a web app with an AI-powered value estimator so that recycling feels as simple as taking a photo. Every scan is a chance to give a device a second life while reducing environmental harm.
What it does
EcoVision lets users scan or upload photos of their device, answer a few quick condition questions, and instantly receive:
- an estimated resale / recycling value,
- options to recycle, give away, or sell,
- and access to responsible partners for disposal.
The core idea: one scan → one estimate → one cleaner planet.
How I built it
- Frontend: Clean, minimal web interface for image upload, device selection, and condition inputs (camera, speakers, battery, charging port, screen cracks, etc.).
- Backend: A value-estimation engine that uses structured device data (model, storage, RAM, damage level, age) and talks to the ML model.
- ML Model: A supervised learning model that maps device specs + condition to an estimated value range. It’s tuned to be realistic rather than “perfect.”
- Database: Stores device models, specs, condition categories, and curated recycler / NGO / refurbisher options.
- Workflow:
User scans or uploads → selects / confirms device → answers quick condition questions → ML model generates a value range → user chooses Recycle / Giveaway / Sell and gets matched to next steps.
Challenges I ran into
- Getting consistent, reliable data for thousands of phone models.
- Designing an estimator that still works when users provide partial or fuzzy information.
- Handling tricky conditions like cracked screens, dead ports, and weak batteries in a simple way.
- Keeping the response time low so the estimate feels instant.
- Communicating clearly that this is an estimate, not a dealer quote, while still building user trust.
Accomplishments that I'm proud of
- Built a working web app + AI model that gives users an actionable value estimate in under a minute.
- Created a clean, guided flow that makes recycling feel effortless instead of like filling a long form.
- Added condition-aware logic (camera, charging port, cracks, etc.) to make estimates feel smarter and more personalized.
- Integrated a clear “Recycle / Giveaway / Sell” decision flow so users always know their next action.
- Prioritized user privacy—images and data stay tightly controlled, and only move forward if the user decides to proceed.
What I learned
- Device recycling is messy: models, damage levels, regional prices, and user expectations all vary widely.
- Simple UX beats complex forms: fewer, smarter questions lead to better completion and less drop-off.
- ML doesn’t have to be fancy to be useful: clean, relevant data improves results more than exotic architectures.
- Users care more about transparency and ranges than a fake illusion of exact accuracy.
- Real impact on environmental problems needs a mix of tech, awareness, and incentives, not just one of the three.
What’s next for EcoVision
- Add computer vision to detect cracks, dents, and missing parts directly from photos.
- Expand beyond phones to laptops, tablets, accessories, batteries, and small appliances.
- Introduce a reward system where users earn points or CO₂-savings badges for every device they recycle.
- Build deeper partnerships with recyclers, refurbishers, and NGOs to increase real-world impact.
- Launch a live impact dashboard showing number of devices recycled and estimated CO₂ saved.
- Develop a full mobile app for instant scanning anywhere, including camera-based model recognition on-device.
Log in or sign up for Devpost to join the conversation.