OpenTreeID: Advancing Community Forestry with Human-Augmented Computer Vision

The proposed research will develop new tools for automated tree species identification that will advance the collaboration of government agencies, citizen groups, and nonprofit organizations that is critical to contemporary community forestry. Proper species identification is particularly important during a tree inventory, since many diseases and infestations are species-specific. Yet the large number of species for which routine identification is required can be extremely problematic for citizen volunteers with minimal horticultural training. OpenTreeID will leverage current advances in machine learning, visual recognition, and computer vision techniques to reduce the amount of volunteer time spent sorting through potential species options in tree keys or field guides while improving the overall accuracy rate of crowdsourced tree data for host agencies. The time saved on routine tree identifications can then be invested in other tree management activities to support a healthy community forest.Healthy street trees slow the accumulation of greenhouse gases, intercept stormwater runoff, improve air quality, reduce noise levels and surface temperatures, create wildlife habitat, increase property values, and provide shade and windbreak that reduce business and household energy consumption. Proper management of the community forest through greater understanding of its composition will help maximize these benefits and improve the quality of life for local citizens.