Species Recognition Models
Wildlife populations are commonly monitored with audio and visual technology, like motion-triggered camera traps or microphones. Because analysing this data manually is labor-intensive, automation of the process using AI computer models is becoming increasingly popular. Addax provides customised machine learning models able to identify species of your choice, allowing you to efficiently manage and interpret large volumes of data.
Applications
Addax can develop custom computer vision models capable of automatically tagging camera trap images. Animals can be detected and identified to species level; people to poachers and non-poachers; vehicles to company and non-company cars—or any other groups you would like to distinguish. Custom identification models are also available for other applications, such as bird vocalisations or gunshot and chainsaw detection and localisation.
Training the Model
For a model to identify a particular species, it must first be trained on sample data of your target species or group, like images or audio recordings. Besides camera traps, it is also possible to develop models for images captured by handheld, drone, underwater or thermal cameras. The amount of data you will need to supply depends on many factors, including the distinctiveness of the target species, the backgrounds and the project setups. In general, the more training data provided, the more reliable the model will be. However, don’t worry if you suspect having too few images. Addax offers several methods to enlarge your dataset. For example, by leveraging existing models to extract information from bulk folders or filtering training data from similar ecological projects. Addax has access to over 16 million camera trap images, covering 850 species and various other taxonomic levels.
Deployment options
These customised AI models can be integrated into existing dataflows, set up for continuous monitoring with real-time notifications, or used through our open-source AI platform, EcoAssist, designed to facilitate image recognition and analysis for camera trap data—a tool developed by Addax Data Science to support open-source projects in nature conservation. This platform runs models locally, so there is no need for an internet connection. Results can take the form of a spreadsheet file, visualised boxes, crops, or collections of images in subfolders based on their detections. Additional features can be tailored to your needs. However, please note that AI is never perfect: Just like human identifiers, you should not expect 100% reliable results.
Reach out if you have any questions regarding the process or how Addax can help maximise efficiency through automated species identification.
