Scientists have developed a non-intrusive method for measuring fluidic properties in a microfluidic chip. Numerous images of flowing droplets are used to train deep neural networks (DNN) to predict the flow rate, which can be done with extreme accuracy and high resolution.