AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Blind Multimodal Quality Assessment of Low-Light Images

Miaohui Wang, Zhuowei Xu, Mai Xu, Weisi Lin

IJCV 2024 · Image generation › Autoregressive and latent generation · approach: Learning-based image quality assessment · 13 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add multimodal self-supervised alignment stage to image-only model (ablation)BMQA RN-50 · MLIQPLCC: 0.7964 -> 0.8567Improved
add multimodal self-supervised alignment stage to text-only model (ablation)BMQA TransF · MLIQPLCC: 0.7504 -> 0.8209Improved
add text modality to supervised stage (image+text instead of image only) (ablation)BMQA RN-50 · MLIQPLCC: 0.8567 -> 0.8614Improved
add feature-alignment pre-training on MS-COCO (ablation)BMQA RN-50 + TransF · MLIQPLCC: 0.8614 -> 0.8989Improved
use feature-alignment pre-training instead of classification pre-training (ablation)BMQA RN-50 + TransF · MLIQPLCC: 0.8707 -> 0.8989Improved
use feature-alignment pre-training instead of restoration pre-training (ablation)BMQA RN-50 + TransF · MLIQPLCC: 0.8753 -> 0.8989Improved
use image modality instead of text modality alone (ablation)BMQA · MLIQPLCC: 0.7504 -> 0.7964Improved
use Transformer text encoder instead of bag-of-wordsBMQA CNXT-B · MLIQPLCC: 0.8822 -> 0.9121Improved
use ConvNeXT-B image encoder instead of VGG-19BMQA TransF · MLIQPLCC: 0.8731 -> 0.9121Improved
replace ResNet-152 with lighter ResNet-18BMQA · MLIQPLCC: 0.8863 to 0.9025No clear effect
use BMQA CNXT-B + TransF instead of best hand-crafted/deep BIQABMQA · MLIQ entire databasePLCC: 0.8596 -> 0.9121Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy