SalFormer360: Transformer-based model enhances saliency estimation for 360-degree videos

Researchers have developed SalFormer360, a new saliency estimation model for 360-degree videos that utilizes a transformer-based architecture. This model combines the SegFormer encoder with a custom decoder and incorporates Viewing Center Bias to better reflect user attention in immersive environments. Experiments on three large benchmark datasets show SalFormer360 significantly outperforms existing state-of-the-art methods, achieving notable improvements in Pearson Correlation Coefficient on datasets like Sport360, PVS-HM, and VR-EyeTracking. AI

IMPACT Enhances the accuracy of saliency estimation in 360-degree videos, potentially improving applications like viewport prediction and content optimization.

RANK_REASON The cluster describes a new academic paper detailing a novel model for saliency estimation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Hugging Face
• Mahmoud Z. A. Wahba
• PVS-HM
• SalFormer360
• SegFormer
• Sport360
• VR-EyeTracking

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →