Automated Artwork Narrative Generation using Deep Learning

Rajput, Devendrakumar Pramod (2025) Automated Artwork Narrative Generation using Deep Learning. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (3MB) | Preview
Preview	PDF (Configuration Manual) Download (803kB) | Preview

Abstract

This research project investigates the integration of computer vision and natural language processing techniques to automatically generate rich, interpretive narratives for visual artworks. The pipeline starts with the characterization of artworks by the EfficientNet B0 model (Tan and Le, 2019), which is a Convolutional Neural Network that combines both accuracy and computational resources. The model has been pre-trained and fine-tuned using a subset of categories from the WikiArt dataset (Wang et al., 2021b). After the genre of an artwork is guessed, a stylistically consistent prompt is formed and fed into a language model to produce a colorful description of the painting. In previous rounds of this project, coherent, context-aware descriptions were generated using a transformer-based language model GPT-2 Large (Radford et al., 2019). In the present system, overall narrative quality and description detail were greatly increased after GPT-4o (OpenAI, 2023), OpenAI’s state-of-the-art multimodal model (OpenAI, 2024) was included. GPT-4o improves expressiveness, stylistic similarity, and emotional subtlety of the generated texts, further closing the gap for interpretation to that of human experts. To promote standardization and interaction, the whole system is published via a web interface made with Flask. The interface enables a user to upload an image and receive an interpretive narrative in seconds, making the tool applicable in educational environments, virtual art galleries, or museum kiosks. The system has been tested to provide good results, though captions from GPT-4o are much more detailed compared to those produced from more basic models such as GPT-2 Large (Zhang et al., 2023).

Item Type:	Thesis (Masters)
Supervisors:	Name Email Khan, Sallar UNSPECIFIED
Uncontrolled Keywords:	Flask; Python; GPT-2 Large; GPT-4o; EfficientNet B0; WikiArt; OpenAI API
Subjects:	N Fine Arts > N Visual arts (General) For photography, see TR Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence P Language and Literature > P Philology. Linguistics > Computational linguistics. Natural language processing Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence > Computer vision Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence > Computer vision
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	03 Jul 2026 09:37
Last Modified:	03 Jul 2026 09:37
URI:	https://norma.ncirl.ie/id/eprint/9452

Actions (login required)

View Item