BrightVQ

The subjective dataset behind BrightRate: Quality Assessment for User-Generated HDR Videos. WACV 2026.

Shreshth Saini, Bowen Chen, Yilin Wang, Neil Birkbeck, Balu Adsumilli, Alan C. Bovik. UT Austin and Google.

Paper (WACV 2026 open access) · Supplementary · GitHub (dataset + BrightRate inference) · BrightRate-LM (follow-up MLLM models)

BrightVQ has 300 HDR user-generated source videos, each transcoded along a bitrate ladder into 2,100 clips in total, rated in a crowdsourced study that collected 73,794 subjective scores. It is a benchmark for no-reference HDR-UGC video quality models. BrightRate, the model in the paper, combines CONTRIQUE features, CLIP semantics, HDR features from a piecewise luminance transform, and temporal differences, and reaches 0.889 SROCC on BrightVQ against 0.853 for HIDRO-VQA.

What is in this repo

File	Rows	Contents
BrightVQ.csv	2,100	One row per clip with MOS, ladder rung, and frame metadata
BrightVQ.txt	2,100	Hashed video ids, one per line, for bulk download

Columns

Column	Meaning
Video	Hashed id, also the S3 object name
mos_j	Mean opinion score (SUREAL)
sos_j	Standard deviation of opinion scores
ref	1 for a source video, 0 for a transcode
name	Ladder rung plus source file name
bitladder	Rung id, e.g. 720p_2M_
resolution	360p, 720p, 1080p, or ref
bitrate	0.2M, 0.5M, 1M, 2M, 3M, or ref
orientation	Portrait or Landscape
framerate	Native playback frame rate
content_name	Source file name
height, width	Frame size in pixels

Loading the scores

from datasets import load_dataset
ds = load_dataset("shreshthsaini/BrightVQ", split="train")

import pandas as pd
df = pd.read_csv("hf://datasets/shreshthsaini/BrightVQ/BrightVQ.csv")

Downloading the videos

Each clip is an HDR MP4 at https://ugchdrmturk.s3.us-east-2.amazonaws.com/videos/<Video>.mp4.

while read v; do aws s3 cp "s3://ugchdrmturk/videos/${v}.mp4" ./BrightVQ_Videos/; done < BrightVQ.txt

License

CC BY-NC-SA 4.0, matching the license file in the GitHub repo. Research use only.

Citation

@InProceedings{Saini_2026_WACV_BrightRate,
  author    = {Saini, Shreshth and Chen, Bowen and Wang, Yilin and Birkbeck, Neil and Adsumilli, Balu and Bovik, Alan C.},
  title     = {BrightRate: Quality Assessment for User-Generated HDR Videos},
  booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  month     = {March},
  year      = {2026},
  pages     = {1522-1532}
}

Contact: Shreshth Saini, saini.2 at utexas.edu.