Object Recognition Speech Recognition Scene

• In our work, we apply deep learning in design engineering (specifically, microfluidic device or lab-on-a-chip design). • Controlling shape and location of a fluid stream enables creation of structured materials, preparing biological samples, and engineering heat and mass transport. • Recent work is capable of manipulating cross sectional fluid shapes by creating chaos by using a sequence of pillars of various diameters and positions in the fluid channel (Fig 1, 2). • Discretizing pillar configuration as integers (Fig 4): The combination of position and diameter of each pillar is assigned an index (or class). A null index is assigned when a pillar is absent at a particular location. This results in 33 classes per pillar in our experiments. • Solving the problem by classification: The input is the desired flow shape (that can be drawn freehand). A convolutional neural network (CNN) extract pertinent features from the input images and simultaneously predicts the index of each pillars in the sequence which produces the desired shape (Fig 5)