Few-Shot Personalized Saliency Prediction Based on Adaptive Image Selection Considering Object and Visual Attention.

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Machine Learning

Background:

• Personalized saliency prediction (PSM) typically requires large datasets.
• Existing methods for finding similar users are burdensome and unrealistic.
• A need exists for efficient PSM with limited training data.

Purpose of the Study:

• To develop a few-shot personalized saliency prediction method.
• To introduce a novel adaptive image selection (AIS) scheme.
• To reduce the data requirements and user burden in personalized saliency prediction.

Main Methods:

• Proposed an adaptive image selection (AIS) scheme.
• Focused on the relationship between human visual attention and image objects.
• AIS selects images based on object diversity and variance in personalized saliency maps (PSMs).

Main Results:

• The proposed method enables effective personalized saliency prediction with few-shot learning.
• AIS successfully selects images with high object diversity and PSM variance.
• Experimental results validate the effectiveness of the new image selection scheme.

Conclusions:

• The novel AIS scheme significantly improves few-shot personalized saliency prediction.
• This approach reduces the burden on users by minimizing the number of required training images.
• The method offers a practical solution for personalized saliency prediction in data-scarce scenarios.