With the increasing demand for entertainment such as animation and games, various games and animation characters have increased. It is time-consuming and laborious to identify specific animation characters manually. In this paper, the system is based on the Python, PyQt5 graphics framework and uses the TensorFlow machine learning framework combined with transfer learning to build a MonaCNN recognition model for image recognition and screening. The experimental results show that the constructed neural network model has an absolute accuracy of 95.1%, which solves the problem that users want to pick out pictures with specific characters from the pictures and save them, and at the same time, exclude pictures without corresponding characters in the picture.