top of page


This project uses some machine learning methods in order to create a simplified cartoon character depending on the input image. 'Cartoonified' results are based on people's traits, such as gender, their hairstyle (including curly or straight, long or short or bold, facial hair), their face shape, whether they wear glasses, and also their body motions. 

The machine learning model used for gender detection, hairstyle detection can be concluded as image classification models. For face shape and glasses, I didn’t use the machine learning model. Instead, I used Python Dlib and the CV library to analyze. I refer to the paper: glasses detection on real images based on robust alignment. I used "tensorflow.js" poseNet model to detect the body motions of the figures. 

Well, the "poseNet" model to detect the pose may cause some problems. If the input images don't include the whole bodies, the leg and the arm positions will be detected wrong, especially legs, they will be randomly created, just as the last image of Shakira shown (Click show more to see the image).

bottom of page