CompressedVQA-AEV: Full-Reference and No-Reference Quality Assessment Models for Asymmetric Encoded Videos
2026-07-06 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMultimedia
AI summaryⓘ
The authors developed two computer models to judge video quality: one that compares a bad video to its original version (full-reference) and one that works without seeing the original (no-reference). The full-reference model uses a special technique called Swin-B to measure differences between videos. The no-reference model analyzes video frames individually using two methods, combines the results, and predicts quality without needing the original. Their full-reference model won first place in a competition, and their no-reference model ranked fourth, showing their methods work well.
Video Quality AssessmentFull-Reference ModelNo-Reference ModelSwin TransformerCompressed VideoTemporal PoolingCross-Fold EnsemblingPerceptual Quality Prediction
Authors
Wei Sun, Xingwei Liu, Dandan Zhu, Xiangyang Zhu, Weixia Zhang, Guangtao Zhai
Abstract
This report presents our solutions to the QoMEX 2026 Grand Challenge on Video Quality Assessment for Asymmetric Encoded Videos, comprising a full-reference (FR) model, CompressedVQA-AEV-FR, and a no-reference (NR) model, CompressedVQA-AEV-NR. The FR approach leverages a Swin-B backbone to extract multi-stage similarity statistics between reference and distorted videos for quality prediction. For the NR setting, our model employs complementary frame-level encoders based on SigLIP2 and Swin-B, followed by temporal mean pooling and cross-fold ensembling to estimate perceptual quality without reference data. Our CompressedVQA-AEV-FR achieves first place in the FR track of QoMEX 2026 Grand Challenge, while CompressedVQA-AEV-NR secures fourth place in the NR track, demonstrating the effectiveness of our proposed models. The code is available at https://github.com/sunwei925/CompressedVQA-AEV.