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

2026-02-12 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Gaming & Interactive Media · Depth: Advanced, medium

Summary

Li-Heng Chen, Alan Bovik, and colleagues introduce MTL-VQA, a multi-task learning framework designed for no-reference video quality assessment (NR-VQA) of gaming videos. This approach addresses the challenges of limited human-rated datasets and the unique visual characteristics of gaming content, such as fast motion, stylized graphics, and compression artifacts. MTL-VQA utilizes full-reference (FR) metrics as supervisory signals, enabling it to learn perceptually meaningful features without requiring human labels for pretraining. By jointly optimizing multiple FR objectives with adaptive task weighting, the framework develops shared representations that effectively transfer to NR-VQA tasks. Experiments conducted on gaming video datasets demonstrate that MTL-VQA achieves performance competitive with existing state-of-the-art NR-VQA methods in both MOS-supervised and label-efficient/self-supervised environments.

Key takeaway

For AI Scientists developing video quality assessment models for gaming content, MTL-VQA offers a robust method to overcome data scarcity. You should consider integrating multi-task learning with full-reference supervisory signals to pretrain models, thereby reducing reliance on expensive human-labeled datasets and improving performance on challenging gaming video characteristics.

Key insights

MTL-VQA uses multi-task learning with FR signals to enable label-efficient NR-VQA for gaming videos.

Principles

• Multi-task learning improves feature transferability.
• FR metrics can supervise perceptual feature learning.

Method

MTL-VQA jointly optimizes multiple full-reference objectives with adaptive task weighting to learn shared perceptual representations, which are then transferred to no-reference video quality assessment.

In practice

• Apply FR metrics for pretraining NR-VQA models.
• Use adaptive weighting for multi-task optimization.

Topics

• No-Reference VQA
• Gaming Video Quality
• Multi-Task Learning
• Perceptual Representations
• Full-Reference Metrics

Code references

• xinyiW915/DIVA-VQA
• LuoweiZhou/VLP
• qyp2000/KVQ

Best for: AI Scientist, Research Scientist, AI Researcher, Machine Learning Engineer, Computer Vision Engineer

Related on AIssential

• ViQ: Text-Aligned Visual Quantized Representations at Any Resolution — ViQ balances semantic richness and detail preservation in discrete visual representations for efficient, unified multimodal modeling.
• PointQ-Bench: Benchmarking Diagnostic and Interpretable Point Cloud Quality Assessment — PointQ-Bench extends point cloud quality assessment beyond scalar scores to diagnostic and interpretable understanding.
• TempGlitch: Evaluating Vision-Language Models for Temporal Glitch Detection in Gameplay Videos — Vision-Language Models struggle significantly with temporal glitch detection in gameplay videos, performing near chance.
• Beyond Scalar Distances: Semantic Attribute Gradients from Frozen MLLMs for Visual Embeddings — SAGA uses frozen MLLMs to provide attribute-resolved supervision for vision encoders, moving beyond scalar distances.
• Analytic Score Optimization for Multi Dimension Video Quality Assessment — Multi-dimensional VQA with interpretable annotations and a novel optimization method improves video quality assessment.
• Representation Learning Enables Scalable Multitask Deep Reinforcement Learning — Scalable multitask reinforcement learning is primarily driven by representation learning, not complex model-based planning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.