Research Article
UltraGlobal: An Enhanced Approach to Image Retrieval Using Global Features
@INPROCEEDINGS{10.4108/eai.21-11-2024.2354612, author={Xuanlang Dai and Pengfei Huang and Shicheng Wang and Zhiqi Zhang and Mingyang Gao}, title={UltraGlobal: An Enhanced Approach to Image Retrieval Using Global Features}, proceedings={Proceedings of the 2nd International Conference on Machine Learning and Automation, CONF-MLA 2024, November 21, 2024, Adana, Turkey}, publisher={EAI}, proceedings_a={CONF-MLA}, year={2025}, month={3}, keywords={image retrieval visual search superglobal global feature deep learning}, doi={10.4108/eai.21-11-2024.2354612} }- Xuanlang Dai
Pengfei Huang
Shicheng Wang
Zhiqi Zhang
Mingyang Gao
Year: 2025
UltraGlobal: An Enhanced Approach to Image Retrieval Using Global Features
CONF-MLA
EAI
DOI: 10.4108/eai.21-11-2024.2354612
Abstract
Image retrieval is an important task in vision, and for a long time, the research has focused on traditional algorithms or DL methods, but neglected the better measurement of feature extraction brought by the combination of the two, based on this, we propose UltraGlobal, an Image retrieval method that uses DL method to extract features and encoding traditional algorithms, and our main contributions are: 1) the introduction of PANet in the feature extraction stage; 2) adopt long Global descriptors and improve GeM pooling; 3) NetVLAD(VLAD) was introduced as an encoding layer; Experimental results demonstrate that UltraGlobal significantly outperforms existing methods on standard benchmarks, showcasing exceptional scalability and precision. This approach offers a more efficient and accurate solution for image retrieval systems. Code: https://github.com/Lennox-Dai/UltraGlobal.
