We present an algorithm for automatic detection of solid- phase (polony) Polymerase Chain Reaction (PCR) objects. The goal is to detect the location and size of each polony present in an image. Using a statistical model for an image of a polony, we are able to weigh different hypothesis, including arrangements of multiple overlapping objects. The algorithm uses a coarse-to-fine approach. A coarse version of the model is used to find candidate object locations and sizes with a low computational cost. More accurate versions of the full model, are then used to test for object presence only at the candidate locations.