🌿 About Verdanta
✨ Inspiration
The idea for Verdanta came from a personal experience. My grandma was gifted a beautiful plant—but she didn’t know what kind it was or how to care for it. With no instructions and no way to identify it, the plant eventually withered and died. She was heartbroken.
That moment really stuck with me. I wanted to build something that could help people like her—people who love plants but don’t always know what they need. Verdanta is a smart, friendly AI plant expert designed to make plant care effortless and accessible for everyone.
🛠️ How We Built It
Although AR functionality is coming soon, Verdanta already allows users to snap a photo of any plant, and then uses ChatGPT-4 to identify it and provide rich care guidance. Here’s how it works:
🔍 Plant Scanning Flow: 1. The user taps “Scan a Plant” 2. The device camera opens and captures an image of the plant 3. The image is converted to Base64 4. The encoded image is sent to GPT-4 with vision capabilities 5. GPT-4 returns the plant’s name, ideal lighting conditions, watering schedule, and care tips
🌱 Greenhouse Management: • Once a plant is identified, users can add it to their greenhouse • They can set watering reminders • Choose the location (bathroom, balcony, greenhouse, etc.) • Track the plant’s health over time with care history
🧠 Tech Stack: • Frontend: React (mobile-first, camera interface) • AI: GPT-4 (image + text) handles both plant ID and conversational care instructions • Backend: Supabase, schema designed with help from GPT-4 to store user plant data, care logs, and notification preferences • Previous Experiments: We initially tried using PlantID APIs but found accuracy and flexibility lacking, so we transitioned fully to GPT-4 for better user experience
🧠 What We Learned • GPT-4’s vision capabilities are incredibly powerful for building human-friendly AI experiences • Simplifying user onboarding and scan flow makes a huge difference for accessibility • Many users don’t want too much data—they just want to feel confident they’re doing the right thing for their plant • Supabase was an effective backend for real-time plant reminders and location tagging
🧗 Challenges We Faced • PlantID API Limitations: We tried integrating external plant ID services but had to abandon them due to low accuracy and poor flexibility • Model Responsiveness: Getting GPT-4 vision to consistently output structured plant data (with care info, taxonomy, etc.) took tuning and prompt iteration • Camera & Upload UX: Ensuring smooth image capture and upload across devices, while keeping performance snappy • Reminder Logic: We had to create a flexible reminder system that supports different watering frequencies, based on the plant type
🌱 What’s Next • 🌐 Augmented Reality Mode: We’re working on an AR interface that overlays plant health info and light-level indicators in real space • 🪴 Community Greenhouse: Users can share their plant journeys, tips, and get help from others • 💬 Voice-based Care Agent: A voice-friendly assistant that gives care advice, like “How do I save my aloe vera?” • 📦 Gift-A-Plant Feature: Let users digitally register gifted plants with notes and care instructions
Verdanta is our tribute to every plant lover who just wants to do better. It’s a tool to bridge the gap between curiosity and confidence—turning every home into a thriving green oasis, powered by AI.
Log in or sign up for Devpost to join the conversation.