Journal Article

PolyMeme: Fine-Grained Internet Meme Sensing

Petrantonakis, Panagiotis C., Arailopoulos, Vasileios, Papadopoulos, Symeon, Koutlis, Christos

2024

Overview

Internet memes are a special type of digital content that is shared through social media. They have recently emerged as a popular new format of media communication. They are often multimodal, combining text with images and aim to express humor, irony, sarcasm, or sometimes convey hatred and misinformation. Automatically detecting memes is important since it enables tracking of social and cultural trends and issues related to the spread of harmful content. While memes can take various forms and belong to different categories, such as image macros, memes with labeled objects, screenshots, memes with text out of the image, and funny images, existing datasets do not account for the diversity of meme formats, styles and content. To bridge this gap, we present the PolyMeme dataset, which comprises approximately 27 K memes from four categories. This was collected from Reddit and a part of it was manually labelled into these categories. Using the manual labels, deep learning networks were trained to classify the unlabelled images with an estimated error rate of 7.35%. The introduced meme dataset in combination with existing datasets of regular images were used to train deep learning networks (ResNet, ViT) on meme detection, exhibiting very high accuracy levels (98% on the test set). In addition, no significant gains were identified from the use of regular images containing text.

• Publisher: MDPI AG
• Subject: Algorithms / Annotations / Datasets / Deep Learning / Hate speech / Humans / Internet / Labeling / meme classification / meme detection / meme taxonomy / Morphology / Neural Networks, Computer / Social Media / Social networks / Taxonomy