This 2D synthetic dataset is composed of 15,248 train and 3,812 test trees. This dataset is intended to both facilitate reproducibility and support development of data-driven tree extraction algorithms. The dataset is generated in two steps: the first step produces the tree geometry, and the second renders the corresponding image. The first step models the angiogenesis process using a simple force-based simulation which iteratively grows a tree composed of a set of nodes and edges. Murray's law is used to assign the vessel diameter to each location along the centerline. Note that the SSA dataset does not contain trifurcations. The second step consumes the tree geometry and produces the noisy rendered image. In order to emulate vessel-like appearance, each vessel is modeled as a 3D cylinder (aligned in the same plane) filled with constant-density contrast. For each tree, a grid of 250x250 pixels is created and rays are traced orthographically through the scene. As a final step in the rendering process, multi-scale Perlin noise is added to the image. More information here: https://github.com/JamesBatten/SimpleSyntheticAngiography