International Conference on Computer and Knowledge Engineering

Home / 14th International Conference on Computer and Knowledge Engineering

A Novel Approach for Image-Text Matching Cross-Modal Space Learning

Authors :

Amirreza Ebrahimi¹ Mohammad Javad Parseh² Pejman Rasti³

1- jahrom university 2- jahrom university 3- Universite d’Angers,

Keywords :

Image-text matching،Computer Vision،NLP،Visual-semantic embedding

Abstract :

Image-text matching, a crucial area of study in image processing and AI, involves computing the similarity between a natural language sentence and an image to create a unified space for comparison. Traditional techniques often struggle to bridge the inherent gap between visual and verbal communication, leading to suboptimal performance. Our approach addresses this challenge by employing advanced matrix operations that directly handle the distinct characteristics of visual and textual data. This innovative method enhances the speed and accuracy of the matching process, reduces computational complexity, and eliminates the need for additional resources. Experimental results demonstrate significant improvements in matching precision and processing time, underscoring the potential of our method to advance the state-of-the-art in image-text matching. This research contributes to the broader field of multimodal AI, paving the way for more integrated and sophisticated systems capable of understanding and interpreting complex visual and textual information. These findings highlight the transformative potential of our approach in advancing the field of image-text matching.