WAON SigLIP 2 Base
WAON SigLIP 2 Base fine-tunes Google's siglip2-base-patch16-256 on LLM-jp's WAON dataset. It is an image-text representation model.
Research release
Japanese cultural images are part of the evaluation
The model card evaluates WAON-Bench, a Japanese cultural image classification benchmark, alongside image-text retrieval and other image classification tasks. It also compares an identically sized model tuned on ReLAION.WAON SigLIP 2 Base official model card
Evaluate local cultural terms directly
Practical guidance. For a Japanese image search system, include regional objects, events and labels in evaluation. General image benchmarks can miss the vocabulary that makes a local catalog useful.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | waon-siglip2-base-patch16-256 |
| Parameters | 375 million.WAON SigLIP 2 Base official model card |
| Image configuration | 16-pixel patches; 256-pixel model variant.WAON SigLIP 2 Base official model card |
| License | Apache 2.0.WAON SigLIP 2 Base official model card |
Related models
- LLM-jp CLIP ViT Base/16
LLM-jp / NII LLMC · LLM-jp CLIP
A 248 million-parameter Japanese image-text encoder for retrieval and zero-shot image classification.
Launch date not confirmedResearch releaseDownloadable - LLM-jp CLIP ViT Large/14
LLM-jp / NII LLMC · LLM-jp CLIP
A 467 million-parameter Japanese image-text encoder for retrieval and zero-shot image classification.
Launch date not confirmedResearch releaseDownloadable