Accelerating Computer-Based Recognition of Fynbos Leaves Using a Graphics Processing Unit

A significant advancement in plant species identification for the Cape Floristic Kingdom (CFK) has been achieved through GPU-accelerated technology. The Fynbos Leaf Optical Recognition Application (FLORA) now offers faster processing speeds without compromising accuracy, with the GPU-accelerated version saving up to 51.85 milliseconds on average for high-resolution images. This development addresses the challenges faced by volunteers and casual workers in managing this biodiverse region, which is threatened by wildfires and invasive species. The study compared three versions of FLORA—MATLAB prototype, C++ non-accelerated, and C++ GPU-accelerated—and demonstrated a notable speed-up of 8.7 times for the C++ version over the MATLAB prototype. This innovation paves the way for more efficient large-scale plant identification tasks, such as rebuilding feature descriptors in leaf databases.

Winberg, S. L., Ramone, M., & Naidoo, K. (2017). Accelerating Computer-Based Recognition of Fynbos Leaves Using a Graphics Processing Unit. South African Computer Journal, 29(3). https://doi.org/10.18489/sacj.v29i3.432