Official Porifera Classifier Computer Vision Project

Traditional classification methods of marine sponges are labour-intensive and depend on expertise, limiting their scalability, among many other complications. To address these challenges, we developed the “Porifera Classifier”, a deep learning, computer vision model trained on a curated dataset of 16,915 labelled images that allows it to detect and classify up to 126 species of marine sponges. Established through a YOLOv8 architecture, the model ensures state-of-the-art accuracy benchmarks based on the detection of amorphous and highly varying structural objects. It further evaluates the integration of algorithms that compensate for light scattering in underwater recordings. The annotated dataset also serves as a valuable resource for future studies. In addition to sponges, the Porifera Classifier dataset includes plankton images to improve the model's robustness and reliability in real-world applications. This addition enables the model to distinguish sponges more accurately within complex underwater environments, reduces misidentification, and enhances its versatility for broader ecosystem monitoring, further extending its value for marine research and conservation efforts.