A Global Precision View for Information Retrieval Evaluation Adapted to Image Retrieval Systems

Authors

Tahar Gherbi
Math and Computer Sciences Dept. Echahid Cheikh Larbi Tebessi University, Tebessa
Ahmed Zeggari
Math and Computer Sciences Dept. Echahid Cheikh Larbi Tebessi University, Tebessa
Zianou Ahmed Seghir
Faculty ST, ICOSI Lab, University of Khenchela, Khenchela

Synopsis

This paper presents a comprehensive analysis of the prevalent evaluation metrics employed in content-based image retrieval. Initially, these widely used metrics are inspired in and influenced by general information retrieval principles, which primarily focused on textual data rather than visual content. In addition, collecting together all or the most of relevant results is not considered by the standard evaluation measures. However, this characteristic is crucial in the context of visual information retrieval. This paper underscores the need for a novel evaluation metric that addresses this particular characteristic.

ICAECE2023
Published
February 5, 2024