AI-Powered System for Automated Image Caption Creation and Recommendations
Dr. Babaso Shinde, Tejaswini Thube, Sneha Nagargoje, Sapna Bhujbal, Khushi Mujawar · Zenodo (CERN European Organization for Nuclear Research) · 2025
With the increasing demand for intelligent automation and visual understanding, Artificial Intelligence (AI)-based image captioning systems have become a vital innovation in content analysis, accessibility, and digital media management. This project presents an AI Image Caption Recommendation System that automatically generates meaningful and context-aware captions for images using deep learning and natural language processing (NLP). The proposed system integrates modern web technologies—HTML, CSS, JavaScript, and Python (Flask)—with TensorFlow/Keras, MySQL, and advanced CNN–LSTM architectures to achieve high accuracy and linguistic relevance in caption generation. The system is composed of several interconnected modules. The Image Upload and Feature Extraction Module utilizes pre- trained Convolutional Neural Networks (CNNs) such as InceptionV3 or VGG16 to extract semantic image features. The Caption Generation Module employs Long Short-Term Memory (LSTM) networks to translate these visual features into descriptive text sequences. The Caption Recommendation Module generates multiple captions and ranks them using BLEU and cosine similarity scores, ensuring users receive the most relevant suggestions. A MySQL Database Management Module securely stores images, extracted features, and generated captions for retrieval and analysis. An Admin Dashboard provides dataset management, performance monitoring, and model retraining functionalities, while the Evaluation Module calculates BLEU, METEOR, and CIDEr scores to assess caption quality. The Flask-based web interface enables seamless interaction between users and the backend model, offering a smooth and interactive caption generation experience. By combining computer vision, deep learning, and NLP, this system bridges the gap between visual and linguistic understanding. It enhances accessibility, improves content management, and provides intelligent caption recommendations applicable to social media, assistive technologies, and automated content labeling.