MegaDetector: Pre-Trained Animal Object Detection for Camera Trap Images
About this resource
MegaDetector is an AI model that identifies animals, people, and vehicles in camera trap images (which also makes it useful for eliminating blank images). This model is trained on several million images from a variety of ecosystems. Introduced in 2018 by Dan Morris, supported by Microsoft, via Sarah Beery in the paper “The MegaDetector: Large-Scale Deployment of Computer Vision for Conservation and Biodiversity Monitoring”, MegaDetector has become a widely used model for animal, human, and vehicle detection in camera trap images. It being specifically trained on camera trap images allows it to excel in wildlife conservation filtering tasks. It is being used in numerous conservation efforts, including at the Wisconsin Department of Natural Resources (WDNR) Snapshot Wisconsin trail camera project.
Key features
- Architecture: MegaDetector v5 is a pre-trained model that relies on the You Only Look Once (YOLO) v5 architecture.
Timeline context
MegaDetector builds on advances in the YOLO family of object detection models and is widely adopted by researchers working with camera trap data.
- YOLOv5 (2021): An updated framework base on the YOLO object detection model.
- YOLO (2015): The initial introduction of the YOLO object detection model.
Model playground
Tutorials and Getting Started Notebooks
- MegaDetector GitHub page: MegaDetector GitHub page has a lot of good information about installation and use.
Questions?
If you have any lingering questions about this resource, please feel free to post to the Nexus Q&A on GitHub. We will improve materials on this website as additional questions come in.
See also
- Microsoft fork - PyTorch Wildlife: Microsoft PyTorch Wildlife is the Microsoft run fork of their support to the MegaDetector project.