In this paper, we propose a CNN-based method to accurately estimate the height of standing children under five from depth images collected using a commercial off-the-shelf smartphone. Overall, we collected data of 3887 children (2581 train data, 1306 test data) aged two-five years in rural India. Our approach estimated height with a mean absolute error of 1.64%, and for 70.3% test images, it achieved the acceptable 1.4 cm range. Hence, our solution can detect stunting accurately, by predicting the estimated height with the child’s age.