FaunaPulse

FaunaPulse

About this app

FaunaPulse turns your Android phone into an AI-powered wildlife camera for ecological field research.

Point your phone's camera at a flower, feeding site, nest entrance, or any area of interest, and FaunaPulse detects, tracks, and photographs visiting animals in real time - fully on-device. After a one-time download of the AI models, no internet connection is needed.

KEY FEATURES
- Real-time AI detection and tracking: every animal that enters the area you watch gets a track ID, and every detection is logged with its time
- Capture without AI: motion-triggered photos, or time-lapse photos or video clips, including a night mode using the phone's flashlight
- AI later: find animals, track IDs and names afterwards in saved photos and videos, also in videos you import (an option for older phones)
- Experimental on-device identification of the recorded animals with the BioCLIP 2 vision model or a fixed-class insect classifier, with a confidence per taxonomic rank
- A draggable region of interest (ROI) to focus monitoring on a specific spot
- Field notes and one GPS position per session
- Session summaries with photos, video playback with the AI's boxes and graphs, plus a Dashboard over all sessions
- Works offline: no account, no ads, no analytics, no data ever leaves your phone

WHO IT'S FOR
FaunaPulse was built for ecological research, especially estimating visitation rates of flower-visiting animals such as pollinators, but it can be configured with different detection models for other wildlife-monitoring use cases.

AI MODELS
No AI model comes inside the app. Tell FaunaPulse what you want to watch (pollinators on flowers, insects on a flat surface, or mammals and birds) and it suggests published models and downloads them once: insect detectors (flat-bug, insectDCT), MegaDetector for common animals, and BioCLIP 2 for names. You can also import your own .tflite detection model. The app shows each model's licence and source.

IMPORTANT: RESEARCH PREVIEW (BETA)
FaunaPulse is an experimental research tool in beta: usable in the field, but bugs are still possible. Validation is ongoing. Detection, tracking and identification accuracy depend on your models, device, target organism and field conditions. Please review your results and validate under your own field conditions before drawing scientific conclusions. See the project's GitHub page for full details and known limitations.

PRIVACY
FaunaPulse collects and transmits nothing. There is no account, no analytics and no tracking. All processing happens on your device, and everything recorded stays in the app's local storage unless you choose to export or share it. Full privacy policy on the project's GitHub page.

OPEN SOURCE
FaunaPulse is free, open-source research software (AGPL-3.0). Source code, documentation and issue reporting are available on GitHub: https://github.com/valentinitnelav/fauna-pulse

What’s new

Version 0.9.0-beta.1
- AI models can be downloaded or imported inside the app: say what you want to watch and download the suggested models once from Hugging Face.
- AI later: find, track and name animals in saved photos and videos, also in imported videos.
- Time-lapse video clips, recording with the screen off, field notes, a Dashboard.
- insectDCT insect classifier; BioCLIP 2 on the graphics processor where possible.
Full details: see CHANGELOG.md on the project's GitHub page.