CoinLens

CoinLens, identify, and value your coin collection in seconds with AI.

Comment

Inspiration

My inspiration comes from a deep personal connection to family history. Both my maternal and paternal grandparents left behind collections of old coins—tangible memories of the past. However, navigating the numismatic market to understand their value was a daunting and frustrating experience.

Every time I visited a market or a dealer, I received vastly different quotes. The lack of transparency made me feel vulnerable to scams, and I constantly worried about being cheated out of the true value of these family heirlooms due to my lack of expertise.

I realized that this problem stems from a massive information gap between experts and ordinary people. When I discovered the advanced vision and reasoning capabilities of Gemini 3, I saw a way to bridge this gap.

I built CoinLens to be an AI-powered "honest broker"—a tool that democratizes expert knowledge, helping people like me identify, understand, and value their coins instantly, ensuring they never have to fear being taken advantage of again.

What it does

CoinLens is an intelligent pocket numismatist:

  1. Instant Identification: Uses Gemini 3 to analyze coin photos, identifying year, country, mint mark, and denomination with high accuracy.
  2. Market Valuation: Provides an estimated price range for circulated and uncirculated conditions based on AI knowledge.
  3. Physical Analysis: Estimates physical traits like diameter, weight, and even thickness (visualized in 3D) to help verify authenticity.
  4. Collection Management: Saves your scan history and organizes coins into a digital collection series by country.

How we built it

  • Frontend: Built with React 19 and Tailwind CSS for a responsive, mobile-first experience.
  • AI Engine: Powered by Google Gemini 3 Flash (via @google/genai SDK). We utilized complex prompt engineering to extract structured JSON data, including visual evidence and physical trait estimation.
  • Storage: We used Dexie.js (IndexedDB wrapper) to implement a secure, offline-first database. Your collection lives on your device, ensuring privacy.
  • Visualization: Implemented pure CSS 3D transformations to visualize coin thickness and dimensions dynamically.

Challenges we ran into

3D Depth Perception from 2D Images: It is difficult for standard OCR to determine a coin's thickness. We solved this by instructing Gemini to infer standard physical specifications based on the identified coin variety and visualizing it with a 3D CSS model.

  • Hallucination Control: Ensuring the AI strictly follows the JSON schema for consistent app data. We utilized Gemini's responseSchema and Type definitions to enforce strict typing.

Accomplishments that we're proud of

Grade Estimation: Implementing a feature to specifically grade the wear and tear of the coin (e.g., G-4 to MS-70).

  • Marketplace Integration: Connecting users directly with reputable auction houses for high-value items.
  • Community Verification: Allowing users to crowdsource opinions on ambiguous coins.

Built With

  • and
  • and-numismatic-knowledge-retrieval.-dexie.js-(indexeddb):-for-persistent
  • beautiful
  • coin
  • collection.
  • component-based-user-interface.-typescript:-to-ensure-type-safety-and-code-robustness.-tailwind-css:-for-rapid
  • consistent
  • for
  • local
  • lucide
  • ocr
  • of
  • offline-capable
  • react-19:-for-building-a-responsive
  • react:
  • storage
  • the
  • ui
  • user's
  • utility-first-styling-and-a-modern-aesthetic.-google-gemini-api-(gemini-3-flash):-the-core-intelligence-engine-for-image-recognition
Share this project:

Updates