The CSIP v1 dataset is a roughly cleaned collection of anime art images designed for zero-shot image classification tasks. It contains diverse images from various anime artists, organized to facilitate style recognition. The dataset was created by author deepghs and last updated on November 17, 2025.
Use Cases
- Train zero-shot image classification models based on the dataset's focus on anime style recognition.
- Benchmark style recognition algorithms based on the collection of diverse images from various anime artists.
- Develop contrastive learning models based on the dataset's pre-training design for style comparison.
Strengths
- Dataset is specifically designed for zero-shot image classification tasks.
- Contains diverse images from various anime artists.
- Has undergone an initial cleaning process.
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 unknown, which may limit suitability assessment.
Provenance
- Source
- huggingface
- Freshness
- Last updated 2025-11-17 18:39:03; freshness should be verified.