furproxy provides a collection of captions for furry-themed images sourced from platforms like e621, CivitAI, and booru sites. The dataset contains approximately 7,500 captions, with at least 70% of the complex scenes being human-reviewed and edited. Captions were generated using Gemini 3 Flash and processed through a pipeline involving multi-crop passes and combination.
Use Cases
- Train image captioning models based on the described AI-generated and human-edited text.
- Benchmark the quality of AI-generated captions against human-reviewed edits for complex scenes.
- Fine-tune text-to-image models using the described pairing of furry art images and descriptive captions.
- Study the pipeline for caption generation involving multi-crop passes and combination as described.
Strengths
- At least 70% of the captions, especially for complex scenes, are human-reviewed and edited.
- Approximately 7,500 captions are available.
- Captions are generated using a described pipeline with Gemini 3 Flash and multi-crop passes.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- Images sourced from e621, CivitAI, and booru sites.
- Collection Method
- Captions generated using Gemini 3 Flash AI model, then processed through a human-review and editing pipeline involving multi-crop passes and combination.
- Freshness
- Last updated 2026-04-29 00:28:44; freshness should be verified.