Alpine Vision AI

An AI-powered Android camera application that identifies alpine flora and fauna using a specialized MobileNetV4 model.

Setup Instructions

1. Model & Data Dependencies

Due to file size constraints, the heavy AI models and dictionary images are not directly included in the repository.

Dictionary Images

To download the images for the offline dictionary (which appear in the app's dictionary screen), run the following script:

cd pipeline
uv run python src/build_dictionary.py

This script will query Wikipedia/DuckDuckGo and download the images into app/app/src/main/assets/dictionary_images/. Note: If you receive a 429 Too Many Requests error, you may be temporarily IP-banned by Wikimedia. Please wait a while before trying again.

AI Model

The .tflite model used for inference should be placed in: app/app/src/main/assets/alpine_classifier_fp16.tflite

2. Building the App

Open the project in Android Studio, connect your physical device, and click Run ('app') (or use ./gradlew installDebug). The app is optimized for real devices and utilizes CameraX for high-quality, zero-latency viewfinder-to-capture synchronization.

Features

• Real-time CameraX Integration: Perfectly synchronized field-of-view (FOV) capture using direct PreviewView.bitmap frame extraction.
• Offline ML Inference: Uses TensorFlow Lite for edge-based classification.
• Interactive Dictionary: Filter, search, and view detailed offline profiles of 282 alpine species.