New video quality assessment models win top spots at QoMEX 2026 challenge

Researchers have developed two models, CompressedVQA-AEV-FR and CompressedVQA-AEV-NR, for assessing the quality of asymmetric encoded videos. The full-reference model, CompressedVQA-AEV-FR, utilizes a Swin-B backbone to compare reference and distorted videos, securing first place in the QoMEX 2026 Grand Challenge's FR track. The no-reference model, CompressedVQA-AEV-NR, employs SigLIP2 and Swin-B encoders for quality estimation without reference data, achieving fourth place in the NR track. AI

IMPACT These models advance the field of video quality assessment, potentially improving video compression and streaming technologies.

RANK_REASON The cluster describes a research paper detailing new models for video quality assessment that achieved high rankings in a competition. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.CV →

• CompressedVQA-AEV
• CompressedVQA-AEV-FR
• CompressedVQA-AEV-NR
• QoMEX 2026 Grand Challenge
• SigLIP2
• Swin Bridge

• paper
• other

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