Disturbing Image Detection Using LMM-Elicited Emotion Embeddings

TL;DR

This paper introduces a novel approach for disturbing image detection by leveraging Large Multimodal Models to extract semantic and emotional features, significantly enhancing classification accuracy.

Contribution

The paper presents a new method that combines semantic descriptions and emotion embeddings from LMMs with CLIP features for improved disturbing image detection.

Findings

Achieved state-of-the-art accuracy on the DID dataset.

Effectively utilizes LMM-elicited emotions for image classification.

Significantly outperforms baseline methods.

Abstract

In this paper we deal with the task of Disturbing Image Detection (DID), exploiting knowledge encoded in Large Multimodal Models (LMMs). Specifically, we propose to exploit LMM knowledge in a two-fold manner: first by extracting generic semantic descriptions, and second by extracting elicited emotions. Subsequently, we use the CLIP's text encoder in order to obtain the text embeddings of both the generic semantic descriptions and LMM-elicited emotions. Finally, we use the aforementioned text embeddings along with the corresponding CLIP's image embeddings for performing the DID task. The proposed method significantly improves the baseline classification accuracy, achieving state-of-the-art performance on the augmented Disturbing Image Detection dataset.