
※ This image recognition software requires AI-based assistance for proper recognition.
Please see the revised specifications and reference below. You may adjust the quote and estimated hours accordingly.
We are developing an Android-based app (not limited to Pokémon) that recognizes various types of trading cards via smartphone camera.
Users can scan cards in rapid succession, and the app provides immediate feedback using sound cues:
✅ “Beep” = successful match → auto proceed
⚠️ “Triple beep” = unclear → marked as Pending
Each pending card is reviewed later on a PC-based admin UI.
The app may also integrate ChatGPT (OpenAI API) to describe or explain card details when requested by the user.
We already have a card image database, so the system only needs to run inference (not training).
App does not need Play Store upload — APK delivery is fine.
Accepted stacks: Flutter, Kotlin, React Native, or HTML5. Web-based fallback is OK.
📦 Deliverables: Android APK + web-based (or local) admin UI
📁 GitHub or Google Drive OK for delivery.
🔄 Updated Flow (After TFLite Inference):
Even if TFLite returns only one result, the app should still:
Send the image to an AI-based service (e.g. ChatGPT)
Ask the AI to confirm the match and give a short card description
Present 1–3 possible cards to the user
After the user selects the correct one, redirect to the detailed page (e.g. https://oursite.com/card?id={card_id})
This additional AI step increases confidence and helps in difficult cases (e.g. foil cards, poor lighting, visual distortions).
📌 Reference:
Reddit – Identifying a Magic The Gathering Card