Conditional GANs for Image Captioning with Sentiments
T. Karayil, A. Irfan, F. Raue, J. Hees, A. Dengel
GANsImage CaptioningSentiment AnalysisDeep Learning
Abstract
We explore the use of conditional Generative Adversarial Networks for generating image captions that incorporate sentiment, combining visual understanding with affective language generation.
Overview
Image captioning traditionally focuses on generating factually accurate descriptions. In this work, we extend the task to include sentiment, generating captions that not only describe what is in an image but also convey a specific emotional tone.
Method
We propose a conditional GAN-based approach that conditions the caption generation process on both visual features and desired sentiment. This allows for controllable generation of affective image captions.