Seabird (Tern) Detection - Africa

Seabirds are found near marine habitats, such as sea and wetlands, due to food availability. They mostly feed on fish and insects. The seabird population is declining at a much faster rate compared to other birds as the coastal region is sensitive to pollution, commercial fisheries, habitat degradation, mineral extraction, human disturbance, etc. Seabirds are also endangered by predatory species from both land and water. Apart from the geography of their habitat, they do not have much ability to defend their nest or protect their young ones. Breeding and laying of eggs happen in open habitats, such as bare ground and open sandy or rocky areas, on coasts and islands with little or no nest material.

The Royal tern and Caspian tern are two of the 350 odd seabird species. These adult terns could be of size 45-60 cm weighing 350-750 gm. Their size puts them in the category of small objects and thus we need very high-resolution imagery to detect them. Recent innovations in drones and AI have enabled us to capture high-resolution imagery over a large geographic area and detect objects of different shapes and sizes. Drones also decrease the disturbance to bird population. Drones are easier to deploy and can perform frequent surveys even after disasters like hurricanes, oil spills, etc. This deep learning model helps automate the task of detecting seabirds (Royal and Caspian terns) from high-resolution aerial imagery. This can help in mapping effective site protection areas for seabirds.

Using the model
Follow the guide to use the model. Before using this model, ensure that the supported deep learning libraries are installed. For more details, check Deep Learning Libraries Installer for ArcGIS.

Fine-tuning the model
This model can be fine-tuned using the Train Deep Learning Model tool. Follow the guide to fine-tune this model.

Input
High resolution RGB imagery (1.0 cm resolution). 

Output
Feature class containing detected seabirds.

Applicable geographies
The model is expected to work well with aerial imagery of West African coast or similar geographies.

Model architecture
This model uses the Mask R-CNN model architecture implemented in ArcGIS API for Python.

Accuracy metrics
This model has an average precision score of 0.76 for seabird.

Training data
The model has been trained on the Aerial Seabirds West Africa.

Limitations