In recent years, health sensing through smartphones has been more and more popular with the improvement of sensors and processing capacity. In this paper, we focus on capturing a fingertip video with a smartphone to monitor blood pressure. Current measurement techniques, however, require intrusive methods or inaccurate measurements. We present a low-cost system that uses smartphone cameras and a light source and our self-built convolutional neural network to measure blood pressure. We recruit 34 volunteers to verify our method. After to-fold cross-validation, the mean absolute errors of our model for systolic and diastolic blood pressure were 4.44 mmHg and 3.68 mmHg, meeting the Association for the Advancement of Medical Instrumentation (AAMI) and British Hypertension Society (BHS) Grade A standards for blood pressure monitors. The method can be applied to every mobile device with a camera and has a wide range of applications.