Phaseless imaging approaches provide a significant advantage for systems where maintaining coherency during the acquisition time is difficult. Here, we demonstrate a phaseless, frequency-diverse, computational imaging system that operates at K-band frequencies (17.5-26.5 GHz). The system consists of a cavity-backed metasurface antenna producing spatially diverse radiation patterns that vary as a function of the driving frequency. The frequency-diverse metasurface antenna can be used to form images at microwave frequencies by collecting measurements at frequencies sampled over the operational bandwidth, obviating the need for either mechanically moving parts or phase-shifting circuits. We show that high-fidelity images can be obtained with the metasurface antenna using only the intensity of the measurements by leveraging a sparse variant of the Wirtinger Flow algorithm. In addition to the hardware simplification achieved by using a frequency-diverse approach, we demonstrate a significant reduction in the number of measurements required to reconstruct a given number of voxels for the phaseless imaging problem. This difference from conventional phase retrieval techniques is achieved by leveraging the sparsity concept, simplifying the complexity of the imaging problem.