PhysarumAI Dataset: Experimental Networks of Physarum polycephalum

This dataset contains a large-scale collection of controlled laboratory experiments documenting transport network formation by the slime mould Physarum polycephalum. The organism is known for its ability to form efficient, adaptive networks for resource distribution, making it a widely used biological model for studying decentralized network formation. All experiments were conducted under standardized conditions using 12×12 cm square Petri dishes filled with non-nutritive agar. Nutrient sources (sterilized oat flakes) were placed at predefined spatial locations on a 16×16 grid, representing network nodes. The plasmodium was introduced at the geometric centre of the dish and allowed to grow for 96 hours, after which the resulting network was photographed. The dataset covers 86 spatial configurations with a total of 850 experimental replicates, enabling analysis of both reproducible patterns and biological variability in network formation. For each experiment, the dataset provides original colour photographs and binary segmentation masks of the resulting networks, as well as structured representations in the form of individual adjacency matrices, a General Matrix (aggregated across replicates), a Probability Matrix (empirical connection probabilities), and a Distance Matrix (Euclidean distances between node positions). This dataset is intended to support research in biological network formation, bio-inspired engineering, graph theory, and data-driven modelling of transport systems. It can be used for comparative studies with classical geometric graphs, statistical analysis of network properties, and as training or evaluation data for computational and machine-learning models inspired by biological systems.