Movies2Scenes: Learning Scene Representations Using Movie Similarities

Abstract

like-this information) to define a measure of movie-similarity which is then used to guide the process of learning a general-purpose scenerepresentation. The illustration shown in this figure presents two similar movies with the same genre along with four non-similar movies with different genres. We automatically select thematically similar scenes from similar movies, and use them in a contrastive-learning setting along with randomly selected scenes from non-similar movies. This allows us to learn a general-purpose scene-representation that achieves state-of-the-art results for a wide variety of downstream tasks related to understanding movie scenes.

State-of-the-Art in Nudity Classification: A Comparative Analysis

“…The model is based on InceptionV3 [22] image embeddings and was trained on a private dataset. A novel method for learning scene representations from movies using a ViT-like video encoder and MLP was proposed in [23]. While the study does not focus specifically on image-based nudity classification, the ViT-like encoder used in the model can be utilized for this purpose.…”