Sign in to view source links and access this dataset
Description
Two image subsets exported from a local dataset folder, containing 600 and 450 rows respectively. The dataset was uploaded by thu-coai and was last updated on June 3, 2026. The syncred_600 subset preserves original annotations, while fp_450 contains metadata derived from filenames.
Use Cases
Benchmarking computer vision models based on the two distinct image subsets.
Training models on image data with embedded annotations, as the image bytes are stored directly in the data files.
Comparing model performance on data with full annotations versus metadata derived from filenames.
Developing or testing data loaders for Parquet files containing embedded image bytes.
Strengths
Contains two distinct subsets with specified row counts: 600 and 450.
The syncred_600 subset preserves original annotations from a meta.json file.
Data files are self-contained with image bytes embedded, removing dependency on external file paths.
Limitations
Description metadata is limited; actual data quality requires manual inspection after download.
Column-level documentation is absent; field semantics must be inferred after download.
Row count is known only for the two subsets; the total scale of the original dataset is unknown.
Provenance
Source
huggingface
Collection Method
Exported from a local dataset folder.
Freshness
Last updated 2026-06-03 09:13:40; freshness should be verified.
License is unknown and must be verified before use.