DroneMegaDetector: AI for detecting wildlife in drone imagery

by

Analysing the large volumes of imagery generated by drone surveys remains a major challenge for wildlife monitoring. In a study led by Chaim Chai Elchik, we developed DroneMegaDetector, a general-purpose AI model designed to detect animals across different species, habitats and drone survey conditions.

Unlike models developed for a particular species or study site, DroneMegaDetector is intended as a tool for aerial wildlife monitoring. It produces bounding boxes around detected animals, creating opportunities not only for counting, but also for subsequent species identification, tracking and other conservation applications

The model and inference code are openly available through GitHub and Hugging Face, and the GitHub version now supports both images and video. Go to the ConservationdronesAI tab to go to our GitHub.

Help us improve DroneMegaDetector

We are looking for more drone images and videos of wildlife to help expand and improve the model. If you have aerial wildlife imagery that you may be willing to contribute, particularly from species, habitats or survey conditions not yet well represented, please get in touch with the Conservation Drones team through conservationdrones@gmail.com.

Key takeaway: Our goal is to build an increasingly useful, open and general AI tool for analysing wildlife imagery collected by drones.

Preprint: Chaim Chai Elchik et al. (2026), DroneMegaDetector, A Generalised Model For Detecting Large Mammals In Aerial Imagery. bioRxiv. https://doi.org/10.64898/2026.10.12.758430

Category : Drone applications