I have developed a deep learning project that predicts the panels to be used by a comic book reader. It is designed for Whakoom, a company that is building the Netflix of comics. Leveraging the work of labelers, we have created a model using over 200k labeled panels, which has significantly increased the efficiency of the labelers from 500 pages per day to over 2000 (and counting).
One interesting aspect of this project was the initial uncertainty about whether the model could capture the unique features of each comic’s style. The panels are labeled based on an art director’s style guide so very subjective decisions..
This project showcases my ability to apply deep learning techniques to improve efficiency and automate processes in the comic book industry.
