Genre Classification of Movies from a Single Poster Image Using Feature Fusion

The movie industry is one of the largest and most influential sectors of any community. Each movie in the industry consists of different elements such as actors, directors, preparation elements, posters, etc. One of the most important elements in any movie is its poster, that plays an important role in attracting the audience. Various information can be obtained from the movie poster, including the movie genre. Today, the movie genre is recognized manually. In this paper, we aim to consider the automatic detection of movie genres based on its poster. Automatic detection of movie genres can have various applications in movie archive systems, search engines, recommender systems, and more. In the proposed method of this paper, four categories of embedding features including the objects in the poster, identifying the actors, age, and gender of the actors in the poster, and their facial expressions are used. Our proposed method is compared with some outstanding previous works over the IMDB dataset poster. By incorporating an ensemble classification approach in our work, the results of our proposed method could achieve the average predicting accuracy of 92% which could outperform the previous works.