Droplet microfluidics, a subset of microfluidics in which water-in-oil emulsions are used as reaction containers for high throughput biological screening protocols. Existing tools to analyze the contents of droplets are based primarily on laser-based flow cytometry, which has high throughput but limited spatial resolution for differentiating the contents of droplets. Visual inspection by microscopy can give the researcher more detailed information about contents, but it is time consuming. Thus, by automating visual inspection using computer machine vision, one can obtain high definition data while also maintaining throughput. Past efforts at developing computer vision-based droplet analysis have limited throughput due to either software constraints (interpreted language -- MATLAB) or hardware constraints (serial processing CPUs). This paper proposes a high throughput droplet analysis algorithm, implemented on the GPU, achieving speeds of ~3000 frames per second, making real-time analysis possible. Real time analysis can be applied to a variety of droplet-based protocols, including polydispersity measurements for digital assays, quantifying cell encapsulation, shape-based chemical detection, and others.