GhostRay: AI-Powered Marine Debris Incentive System

GhostRay is a "Scan-to-Earn" prototype designed to incentivize the collection of marine debris by bridging AI Computer Vision with the Solana blockchain.

🌊 The Vision: Why Blockchain?

Marine debris collection often lacks transparency and immediate rewards. We use blockchain for three core reasons:

• Transparency: Every payout from the "UN Treasury" is visible on-chain, ensuring funds go directly to collectors without middlemen.
• Immutability: Data regarding the mass and type of waste is stored on IPFS, creating a permanent audit trail of environmental impact.
• Instant Incentives: Traditional grants take months; GhostRay delivers borderless SOL rewards in seconds.

🛠️ How It Works (The Core Logic)

1. Verification (The Vision Oracle)

• Object Detection: We use a custom YOLO model to identify debris in real-time.
• Material & Mass Analysis: Captured frames are sent to Google Gemini 1.5 Flash. Gemini acts as an "Environmental Oracle," identifying the material and estimating the mass (kg) of the debris based on its volume and type in the image.

2. The Reward Engine

Rewards are calculated based on ecological impact:

• Mass-Based: The base reward is tied directly to the estimated weight (kg) provided by Gemini.
• Location Multiplier: We use a SustainabilityEngine that adjusts rewards based on the country's environmental index.
• Current Multiplier (Canada): Since the demo is currently running in Canada (CAN), the engine applies a specific multiplier from our sustainability dataset to reflect local economic and environmental standards.

3. On-Chain Settlement

• IPFS Metadata: All scan data is pushed to IPFS to secure the evidence.
• Atomic Payout: Once verified, the backend signs a Solana transaction to send a SOL reward directly to the collector's wallet.

🚧 Project Status & Pivots (Honest Disclosure)

During development, we faced a steep learning curve regarding Metaplex Bubblegum (the protocol for compressed NFTs).

• The Challenge: Implementing cNFTs in Python is technically complex due to Merkle Tree serialization requirements, and we encountered API rate-limits during high-frequency testing.
• The Pivot: To ensure a stable demo, we prioritized the AI-to-SOL Bridge. We focused on making the payment and mass estimation 100% functional rather than shipping a buggy NFT feature.

🚀 Roadmap

• Full Lifecycle cNFTs: Transition to tracking waste from Collected → Verified → Recycled using on-chain state updates.
• Hybrid Architecture: Implement a Node.js microservice specifically to handle high-speed cNFT minting.
• On-Chain Logic: Migrate payout rules from the backend to an Anchor (Rust) smart contract for full decentralization.

💻 Tech Stack

• AI: Ultralytics YOLO, Google Gemini 1.5 Flash.
• Backend: Python (FastAPI), Uvicorn.
• Blockchain: Solana (Devnet), Shyft API, solders library.
• Frontend: Next.js (Tailwind CSS).