How phantombeast7/birdrecon-a-free-open-source-tool-for-image-based-bird-species-recognition works
Main Components
• Mobile Application (Android):
• Located in the “App/app” directory. It’s an Android app built with Android Studio and written in Kotlin.
• Contains various UI components (Home, FAQ, Feedback, Search) and ViewModels within the “birds/species/birdsspeciesclassification/ui” subdirectories.
• It interacts with external endpoints (for both name-based and image-based search) as indicated in the README.
• Backend Services (Server):
• Under “Server File,” there are two main submodules:
▪ “Search By Image”: Contains a Dockerfile, Python script (main.py), ensemble TensorFlow Lite model (ensemble_final.tflite), and CSV indices. This endpoint provides image recognition services.
▪ “Search By name of the Bird”: Contains a Dockerfile and app.py to serve name-based searches.
• Website (Web Application):
• Located in the “Website” directory. This is a Flask-based web app (app.py) with templated HTML pages (index.html, results.html, teams.html) and static assets.
• Acts as an alternate front-end for users to interact with the service.
• External Integrations and Cloud Services:
• Google Gemini API: Used for fetching detailed species information.
• Wikimedia Commons: Used for retrieving related bird images.
• Microsoft Azure: Deployed via Docker containers and Azure CLI commands as outlined in the README.
• Docker: For containerizing the backend services and the website.
Relationships and Interactions
• Both Mobile and Website Clients:
• They are interfaced with the backend APIs. The mobile app specifically references updating endpoint URLs for both image and name searches.
• Data flows from these front-ends to the relevant backend services via HTTP requests.
• Backend API Services:
• The “Search By Image” service leverages machine learning models (via TensorFlow Lite) to classify bird species based on images.
• The “Search By name of the Bird” service handles textual input and returns search results based on bird names.
• Both services may interact with external APIs (Google Gemini and Wikimedia Commons) to enrich the data with species details and images.
• Deployment and Container Orchestration:
• The Dockerfiles in the Server File and Website directories show that these components are containerized.
• Azure is used to deploy these containers (via Azure Container Registry, App Service Plan, and Web Apps).
Architectural Patterns
• Separation of Concerns: The project cleanly separates user interfaces (mobile app and website) from backend services (image-based and text-based search endpoints).
• Microservices-like Structure: Even though the backend is divided into separate directories, each service (image recognition and name search) can be deployed independently in containers.
• API-driven Design: RESTful APIs (evidenced by the endpoints in the mobile code and Python scripts) serve as the contract between frontends and services.
Relevant Technologies, Frameworks, and Libraries
• Android (Kotlin, Android Studio)
• TensorFlow Lite (for image recognition)
• Flask (Python web framework used in both backend and Website)
• Docker (for containerization)
• Microsoft Azure (for deployment and container orchestration)
• External APIs (Google Gemini API and Wikimedia Commons)
Specific Elements
• Mobile App (Android):
• UI screens (Home, Search, FAQ, Feedback)
• Website (Flask Frontend):
• UI Pages (index, results, teams)
• Server Backends:
• “Search By Image” Service:
▪ A block that shows the TensorFlow Lite model (ensemble_final.tflite) processing images.
▪ Integration with external APIs (Google Gemini for detailed species information).
• “Search By Name” Service:
▪ A block that handles textual bird name queries.
• External Dependencies:
• Google Gemini API (detailed species info)
• Wikimedia Commons (similar images)
• Deployment & Infrastructure:
• Communication Protocols: