DeepFaune New England (DFNE) model for species classification analyzes trail camera imagery. This model is a re-trained version of the DeepFaune model for classifying European species in trial cameras, fine-tuned to classify taxa from northeastern North America. DFNE classifies 24 taxa, including the “no-species” label indicating the absence of an animal.

SpeciesNet is runs two AI models: (1) an object detector that finds objects of interest in wildlife camera images, and (2) an image classifier that classifies those objects to the species level. This ensemble is used for species recognition in the Wildlife Insights platform. The code is now open source, allowing AMBER to include SpeciesNet into our machine learning repertoire.

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.

MegaDetector only finds animals, it doesn’t identify them to species level.

BirdNET uses deep learning to recognize over 6,000 species globally within audio files. It was developed by the K. Lisa Yang Center for Conservation Bioacoustics at the Cornell Lab of Ornithology in collaboration with Chemnitz University of Technology.

Go to https://birdnet.cornell.edu to learn more about the project.

HawkEars is a desktop program that scans audio recordings for bird sounds and generates Audacity label files. It is inspired by BirdNET, and intended as an improved productivity tool for analyzing field recordings. AMBER runs HawkEars directly and stores results in your database, accessed through our web portal.

Ruffed Grouse are a flagship species in North America, of interest to many conservation groups. See this article for more details.

When acoustic targets have a recognizable signal, simple template-matching may be a powerful way to screen audio files for a target wildlife species of interest. AMBER has the ability to create customized templates and screen files for potential signal matches, and further develop machine learning models to locate true detections.

ABGQI is a CNN focused on soundscape analysis, predicting 5 soundscape components: human noise (Anthropophony), wildlife vocalizations (Biophony), weather phenomena (Geophony), Quiet periods (Q), and microphone Interference (I). AMBER users may elect to use this model for soundscape monitoring, among others.