Sign in to view source links and access this dataset
Description
The RealQA dataset is proposed for User Generated Content (UGC) assessment, building upon established real-world image quality datasets like AVA, TAD66K, Koniq-10k, LIVE-C, and SPAQ. It contains image data and conversation templates for training and testing. Specific row counts, column counts, and file sizes are not provided in the input.
Use Cases
Train models for Image Quality Assessment (IQA) using real-world image data from sources like AVA and Koniq-10k.
Conduct Image Aesthetic Assessment (IAA) on user-generated content (UGC) leveraging the proposed RealQA dataset.
Utilize provided conversation templates for training and testing dialogue-based assessment systems related to image quality.
Strengths
Builds upon multiple established real-world image quality datasets (AVA, TAD66K, Koniq-10k, LIVE-C, SPAQ).
Specifically proposed for assessing User Generated Content (UGC).
Includes training and test conversation templates for structured evaluation.
Limitations
The dataset size, number of rows, columns, and specific file formats are unknown.
The exact composition and licensing terms for the aggregated source datasets are not detailed.
Sample data is unavailable for preview, making initial assessment difficult.
Provenance
Source
Hugging Face repository by MingxingLi, aggregating from multiple public datasets.
Collection Method
Proposed for UGC assessment, utilizing established IQA/IAA datasets captured in the real world.
Freshness
Last updated on 2025-05-29.
The dataset is distributed as a multi-volume compressed package (dataset_split.zip); the unzip program must support this format. Users should review the full description on the Hugging Face page for complete details. License information is unknown.