Abstract: A cinema population distribution statistical model based on Convolutional Neural Network (CNN) is formulated by detecting facial extraction features to predict people's age and gender based on traditional algorithms. The model includes two parts: face detection model and convolution neural network model. It classifies people accurately according to age and gender, and then gets the distribution of cinema crowd, which provides a reliable basis for cinema film projection. The experimental results on adience open data set show that the accuracy of the designed model is as high as 77.92% for cinema crowd distribution statistics.