žAIba

Introduction

Each spring, thousands of amphibians cross roads on their way to breeding sites. As infrastructure expands, more roads intersect their natural migration paths, resulting in countless frog fatalities due to traffic. While some conservation groups build temporary fences and manually relocate frogs, current systems rely heavily on word-of-mouth, Facebook groups, and manual data collection. There's no consistent, real-time system in place.

Our Solution

žAIba is a real-time, AI-driven support system for volunteers who help amphibians during migration. It combines edge machine learning, LoRaWAN communication, and a mobile app that shows real-time frog activity.

Description

Every spring, thousands of amphibians cross roads on their way to mating ponds, often resulting in mass fatalities due to traffic. Existing protection efforts are based on temporary fences and manual frog transfers – but lack real-time data, coordination, and species-specific insights.

That’s where žAIba steps in.

This project empowers volunteers with real-time notifications about frog presence on specific migration points using AI and LoRaWAN. It combines environmental awareness with cutting-edge technology to make a real-world impact.

Machine Learning with Edge Impulse

We developed two models:

• Detect the presence of frog calls in the environment
• Classify four common Slovenian frog species:
• Hyla arborea (European tree frog)
• Bufotes viridis (European green toad)
• Rana temporaria (European common frog)
• Bufo bufo (European toad)

We contacted the Slovenian Museum of Natural History and searched the Internet for recordings of mentioned frog calls. We also gathered audio of other sounds common to locations of interest, namely forests.

Steps to our ML model:

• Gathered audio
• Labeled and uploaded data to Edge Impulse
• Extracted spectral features
• Trained and validated a neural network

📌 The model runs directly on the NatureGuard Seeed Studio using its onboard microphone.

We deployed the model as an Arduino library.

Connectivity

LoRaWAN Connection

The Arduino device sends frog detection alerts over the LoRaWAN network, enabling remote locations without relying on WiFi or mobile signal. This ensures energy-efficient, long-range data transmission from roadside devices to the central system.

Custom 3D case for sensor module

We modeled a custom sensor enclosure for the NatureGuard board which keeps the device safe from environment factors such as rain and wind.

Native Android App

The app displays:

• Real-time statistics of frog activity by location
• Informative content about species
• Historical migration data
• Campaign coordination tools for volunteers

📹 Demo Video

See how žAIba works in action — from call detection to app notification.

How it all comes together or Do it yourself

1. Arduino device detects frog calls using the trained ML model

2. Detection is sent via LoRaWAN

3. The app receives and displays data for volunteers

4. Action can be taken immediately on the ground

Improvements for the future

• Add support for more species
• Improve battery life and waterproofing
• BLE pairing mode for configuration
• Integration with public campaign maps
• Collaborate with local municipalities

Conclusion

žAIba demonstrates how modern technology — from edge machine learning to LoRaWAN and mobile apps — can support real-world environmental efforts. By enabling real-time frog call detection and species classification, we’ve created a tool that empowers volunteers to act precisely when and where help is needed.

This project is more than a prototype — it’s a vision of how IoT and AI can intersect with nature conservation. With further development, we aim to expand the system, engage wider communities, and make meaningful contributions to amphibian protection across Slovenia and beyond.

Whether you're a developer, biologist, volunteer, or tech enthusiast — we invite you to JOIN US, test our tools, and help build the future of wildlife conservation.

🎓 Acknowledgments

The initiative for the project came from asist. mag. Luka Mali provided us with the proper equipment and knowledge to bring this idea to life. The project was part of faculty subject Mobilnost in internet stvari, part of Multimedia program at Faculty of Electrical Engineering of University of Ljubljana.

Special thanks to dr. Tomi Trilar, museum councillor at the Slovenian Museum of Natural History, for providing authentic frog recordings.

👨‍💻 Team & Credits

Pika Križnar

• Idea development
• Project management
• ML model engineer
• Arduino software development
• Mobile app developer
• Content collection
• UX/UI design

Nina Viktoria Baškarad

• Arduino software development
• Project coordination
• Demo video animation
• Content creator

Gašper Vrabič

• LoRaWAN integration
• Arduino software development
• Prototype configuration
• Technical support
• Content creator

Neža Zore Pokorn

• ML model preparation
• Content collection
• Sensor enclosure design
• Content creator