The challenge faced by the Monarch Joint Venture was multifaceted. The conservation of the monarch butterfly, a key pollinator species, was primarily critical due to its diminishing population. Central to this conservation effort was monitoring milkweed plants, which are essential for monarch reproduction.
Understanding the monarch challenge
The traditional monitoring methods involve manually counting milkweed plants, which is costly, time-consuming, and labour-intensive. These methods required specialized training, producing subjective results through different interpretations of the physical observations.
The project's scale, spanning from Mexico to Canada, demanded a solution for efficient and accurate coverage over vast areas.
Greg Emerick, CEO at SBA, driven by a passion for technology and innovation, sought a partner with technical expertise and shared corporate ethics. After exploring several options, SBA found a perfect fit with Radix.
Aligning values and vision
Radix's in-depth understanding of the problem before proposing a solution resonated with SBA's philosophy. "Finding a partner like Radix, who not only had the technical expertise but also shared our vision for environmental stewardship, was crucial," states Greg Emerick, CEO of SBA. The alignment between Radix’s technical team and SBA's business objectives was key in forging this partnership.
The work initiated by Radix and its partners was systematic and data-driven. The first step involved critically analyzing existing data. This data, primarily aerial images, was initially annotated by non-experts, leading to inconsistencies.
A Collaborative and Innovative Approach
Radix, in collaboration with the Monarch Joint Venture (MJV), undertook a process of re-labeling and correcting this data. Victor Hutse, Solution Lead at Radix, highlights the collaborative nature of the work: "It was a fusion of expertise. Our team at Radix brought advanced AI skills, while MJV contributed invaluable ecological knowledge. This synergy was key to developing an effective solution."
The Model - A Technical Look
Our solution revolves around a Convolutional Neural Network (CNN) object detection model designed to recognize milkweed stems in drone-captured images. We meticulously compared the model's predictions with on-site field counts conducted along designated transects to ensure optimal performance. The result is a highly proficient model, accurately identifying milkweed stems and achieving an impressive mean Average Precision (mAP) score of 0.918—a performance level comparable to that of a human domain expert.
POLLi's goal is to map North America's milkweed and add 1.8 billion stems, crucial for Monarch reproduction. Radix’s AI model is key, accurately identifying milkweed and guiding informed decisions for researchers and policymakers.
The potential applications of POLLi extend beyond the initial project. The success of POLLi in efficiently monitoring and managing milkweed populations opens avenues for its application in other conservation efforts.
Harnessing Technology for Broader Environmental Goals
This form of object detection can be adapted to monitor different species of flora and fauna. "What started as a project to save monarch butterflies has the potential to transform how we approach conservation globally. POLLi is a testament to technology being harnessed for the greater good," reflects Greg Emerick. Furthermore, the underlying technology of POLLi holds promise for various industries concerned with habitat management and reporting, demonstrating their contribution to sustainability and better management practices.
Radix's ongoing AI contribution to POLLi exemplifies how technological innovation, aligned with conservation goals, can lead to transformative environmental management and sustainability practices.