Convolutional Neural Network based Detection and Measurement for Microfluidic Droplets

Modern microfluidic systems realize the envisioned idea to perform continuous process operations on a small scale using miniaturized devices and present superiorities in terms of plant modularization, reaction intensification and waste reduction. In microfluidic engineering, droplet size is central to desired function. Therefore, an effective droplet detection and size measurement method is highly-demand to quantitatively reveal the relationship between operation parameters and outcome droplet size. Herein, with recent impressive developments of computer vision, we propose a novel two-step convolutional neural network method to detect and measure droplets in microscopic images. The proposed model first locates droplets with bounding boxes and then calculate the droplet size with detailed coordinates. This convolutional neural network model not only exhibits outstanding performance for droplet size measurement, but also reveals the convenience of deep learning for digital, comprehensive and intelligent microfluidic researches.