Global-Local Multiple Granularity Learning for Cross-Modality Visible-Infrared Person Reidentification

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Visible-infrared person re-identification (VI-ReID) is crucial for smart surveillance but faces challenges due to cross-modality discrepancies and intraclass variations.
• Existing VI-ReID methods often struggle with effective optimization objectives for learning discriminative features.

Purpose of the Study:

• To propose a novel Global-Local Multichannel (GLMC) network for VI-ReID.
• To introduce an effective center loss function to address cross-modality discrepancies and improve feature learning.

Main Methods:

• The GLMC network learns multigranularity representations by combining global and local features for enhanced feature descriptors.
• A novel center loss function is proposed to simultaneously increase intraclass cross-modality similarity and interclass discrepancy.
• The method explicitly handles cross-modality discrepancies and model fluctuation issues.

Main Results:

• Experimental results on two public datasets demonstrate the superiority of the proposed GLMC network and center loss.
• The method achieves state-of-the-art effectiveness in visible-infrared person re-identification tasks.

Conclusions:

• The proposed GLMC network and center loss function offer a significant advancement in VI-ReID.
• This approach effectively tackles the challenges of cross-modality discrepancy and intraclass variations in person re-identification.