Learning Perceptual Representations for Gaming NR-VQA with Multi-Task FR Signals

2026-02-12

AI Summary

This work addresses the challenges of no-reference video quality assessment (NR-VQA) for game videos, which stem from the absence of human annotations and the inherent complexity of such content—including rapid motion, stylized graphics, and compression artifacts. To tackle these issues, the authors propose MTL-VQA, a multi-task learning framework that innovatively leverages full-reference (FR) quality metrics as self-supervised signals to pretrain perceptual features without requiring human labels. By adaptively weighting and jointly optimizing multiple FR objectives, the model learns transferable shared representations suitable for NR-VQA. Experimental results demonstrate that the proposed method achieves performance on par with state-of-the-art NR-VQA approaches across various settings—including MOS-supervised, label-efficient, and fully self-supervised scenarios—on game video datasets.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

No-reference video quality assessment (NR-VQA) for gaming videos is challenging due to limited human-rated datasets and unique content characteristics including fast motion, stylized graphics, and compression artifacts. We present MTL-VQA, a multi-task learning framework that uses full-reference metrics as supervisory signals to learn perceptually meaningful features without human labels for pretraining. By jointly optimizing multiple full-reference (FR) objectives with adaptive task weighting, our approach learns shared representations that transfer effectively to NR-VQA. Experiments on gaming video datasets show MTL-VQA achieves performance competitive with state-of-the-art NR-VQA methods across both MOS-supervised and label-efficient/self-supervised settings.

Problem

• no-reference video quality assessment
• gaming videos
• perceptual representation
• full-reference metrics
• multi-task learning

Innovation

• multi-task learning
• no-reference VQA
• full-reference supervision
• perceptual representation
• gaming video quality