This paper introduces DVL2024, an expansive ultra-high definition (UHD) video dataset meticulously curated for advancing perceptual video quality assessment (VQA) research.DVL2024 stands as the largest available UHD video quality assessment dataset in stunning 4K resolution ( $3840 imes 2160$ ), making it a highly valuable resource for the research community.The dataset comprises a diverse collection of 643 4K UHD video sequences, each spanning 10 seconds and captured at a high frame rate of 50 frames per second (fps).The videos are stored in raw 10-bit 4:2:0 YUV format, ensuring their fidelity and authenticity.
To ensure the reliability of subjective evaluations, our annotation process adheres to the rigorous guidelines set forth by ITU-R BT.500-13, involving the participation of 32 carefully selected individuals with a wide HEMP SEED OIL age range, spanning from teenagers to sixties individuals, with an average age of 27.2 years.DVL2024 exhibits several notable advantages that contribute to its importance in research.
The dataset encompasses a wide range of video content, capturing diverse scenes and subjects while incorporating recordings from various camera types.It also incorporates a rich variety of complex and realistic distortions, providing a comprehensive evaluation of perceptual quality.The subjective scores provided by the participants are meticulously balanced, resulting in a robust and reliable Mean Opinion Score (MOS) distribution.Furthermore, we conducted benchmark experiments using DVL2024 to assess the performance of widely recognized VQA methods.
The meticulously crafted dataset we have developed has Wearable Hooded Blanket demonstrated exceptional performance in rigorous benchmark evaluations, consistently exceeding the SROCC threshold of 0.7 and even achieving remarkable levels of 0.8 across all tested networks.These findings validate the significance of DVL2024 and pave the way for the advancement of UHD VQA research.
We are committed to making the DVL2024 dataset widely accessible, fostering research and development in video quality assessment.DVL2024 is publicly available at https://github.com/GZHU-DVL/DVL2021.