Social Anchor-Unit Graph Regularized Tensor Completion for Large-Scale Image Retagging.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Science

Background:

• Social image retagging enhances tag quality by correcting and adding tags.
• Current methods using visual, user, and tag data are computationally expensive for large-scale datasets.
• Anchor graphs accelerate large-scale graph learning by focusing on a subset of anchor points.

Purpose of the Study:

• To propose a novel and efficient method for social image retagging that is scalable to large datasets.
• To introduce a multi-domain anchor-unit graph construction for improved tag refinement.
• To leverage tensor completion with a novel regularization for efficient tag assignment.

Main Methods:

• Developed a Social anchor-Unit GrAph Regularized Tensor Completion (SUGAR-TC) method.
• Constructed an anchor-unit graph across multiple domains (image, user).
• Applied tensor completion with Social anchor-Unit GrAph Regularization (SUGAR) for anchor image tag refinement and efficient non-anchor image tag assignment.

Main Results:

• SUGAR-TC demonstrates effectiveness and efficiency in refining social image tags.
• The proposed method is insensitive to the scale of the data.
• Experimental results show superior performance compared to state-of-the-art methods on a real-world social image database.

Conclusions:

• SUGAR-TC offers an efficient and scalable solution for social image retagging.
• The multi-domain anchor-unit graph approach enhances tag refinement accuracy.
• The method significantly improves tag quality for large-scale social image datasets.