POSE: Phase I: Wildbook: Building an Open Source Community for AI-Enabled Wildlife Science and Computer Science Education

Wild Me, Portland OR

Investigators

Abstract

Digital images and video have become the most ubiquitous and inexpensive data sources for wildlife research, especially when well designed scientific efforts can partner with the public to increase the breadth of coverage and volume of data. From population estimation to mapping migration routes and complex social networks, collected imagery of individual animals can foster new discoveries in statistical modeling, AI for computer vision, and conservation biology, as well as provide a data-driven basis for effective resource management policy. The open source Wild Me ecosystem (https://github.com/wildmeorg) aids wildlife researchers at universities and local NGOs in curating large volumes of this visual data, employing a multistage machine learning pipeline to find, count, and even individually identify wildlife in photographs in support of population biology, social ecology, and more. The project’s software-supported data and use cases have advanced university education in computer science and applied AI using compelling, real world data of wildlife to challenge students. Through open source community building, the project can scalably grow its species coverage and its interdisciplinary research impact, providing an advanced foundation for education and wildlife research. This project will advance the Wild Me ecosystem through building an open source managing organization, which is responsible for the initial open source community growth effort (i.e. attracting software professionals and scientists). The managing organization will establish the foundational contribution model for code, AI models, and related data and define the community governance model for the Wild Me ecosystem. The project’s efforts also include developing training materials to encourage open source code contribution, reaching out to related professional and academic users (potential code contributors), and defining the flow of data and machine learning models back and forth between academia and field biologists for interdisciplinary collaboration. The project begins with identifying existing, unmet needs and open source appetite in our existing user community. It then will create a gap analysis translated into needed services and provide code examples and documentation to make it easy and secure for third party contributions. The project will establish quality and security standards for code and model contribution and review, as well as set a code of conduct for community discussion and contribution. The project ultimately establishes the foundation for a thriving and growing open source ecosystem that supports multidisciplinary, collaborative wildlife biology and ecology using AI as an effective tool to scale and speed each research effort. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.