AI-powered environmental monitoring system for marine and terrestrial ecosystems using computer vision, real-time analysis, and FAIR data principles.

Rust library for BirdNET and Perch ONNX model inference with CUDA GPU support

CAMCALT (Complex Animal Movement Capture and Live Transmission) is a forest surveillance and monitoring system designed to capture complex animal movements and provide live video feed wirelessly from any part of the world. It aims to prevent hunting and poaching, enhancing forest security.

Desktop GUI for birda bird species detection — Electron + Svelte 5 interface for BirdNET-powered audio analysis, detection browsing, and species visualization

Userspace programs and source code for the Audiograb system

A python based GUI that runs an onnx detection model over trail camera images. The output can then be manually reviewed image by image to correct any detection errors to give an accurate output for either reporting or for feeding more images back into the detection model for improving accuracy.

CV pipeline for bear detection in wildlife camera traps — YOLOv8 detection & segmentation, SORT multi-object tracking, MobileNetV2 classification. Fine-tuned on 1,172 images · mAP@0.5 0.96

An R package to explore and process camera trap data and automate reporting for Wildlife Monitoring.

A Model Context Protocol (MCP) layer for bioacoustic analysis that augments BirdNET detections with contextual reasoning, uncertainty-aware summaries, and ecologist-oriented insight extraction.

SQL Scripts to automatically populate GeoNature synthesis from VisioNature datas

Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30

Bio-inspired ornithopter robot for wildlife conservation monitoring. Features bio-mimetic flapping mechanism, aerodynamic wing design, MPU-based stabilization and RF control system. Includes CFD analysis, mechanical design, embedded systems and experimental validation.

A commandline tool for finding animals in pictures

Open source tools for wilderness conservation. Built from the granite walls of Cochamó, Chile

Automatic bird species recognition from audio — mel spectrogram + EfficientNet-B0. 20 alpine species, 96.5% accuracy. Xeno-canto API · librosa · MLflow · Gradio app

Real-time AI monitoring for bird nests using camera snapshots, vision models, and Discord alerts

Hanging in there: the highs and lows of a flying-fox camp in a novel landscape

Combating Human wildlife conflict using AI/ML

A wildlife monitoring system using EfficientNet-B3 and XAI (GradCAM) for species classification, detailed HTML/PDF reports, safety protocols, markdown insights, and CrewAI integration for advanced analysis.

WISDAM (Wildlife Imagery Survey – Detection and Mapping) is a software package used for the digitisation of objects within images and to enrich objects with meta-data.