Meet the team that turns your ideas into work.

Discover EightBots
EIGHTBOTS
Contact salesSign up
rinnaJapanese image-text encoders

Japanese CLOOB ViT-B/16

Japanese CLOOB ViT-B/16 learns shared image and Japanese-text representations through Contrastive Leave One Out Boost training.

Japanese captions train the shared space

The training set is CC12M with captions translated into Japanese. The published release date is 12 May 2022. The weights are encoders for image-text comparison, not an image generator.Japanese CLOOB ViT-B/16 official model card

Evaluate retrieval with Japanese queries

Practical guidance. Compare this release with Japanese CLIP using the same image collection and queries. The encoders have similar shapes, but a different training objective can change retrieval rankings.

Specifications & access

Japanese CLOOB ViT-B/16 specifications
DetailValue and conditions
Launch date
AccessDownloadable
Model identifiersjapanese-cloob-vit-b-16
ArchitectureViT-B/16 image encoder with a 12-layer BERT text encoder.Japanese CLOOB ViT-B/16 official model card
AncestryThe image encoder starts from AugReg vit-base-patch16-224.Japanese CLOOB ViT-B/16 official model card
LicenseApache 2.0.Japanese CLOOB ViT-B/16 official model card

Related models

  • Japanese CLIP ViT-B/16

    rinna ยท Japanese image-text encoders

    Japanese CLIP ViT-B/16 learns shared image and Japanese-text representations through contrastive image-text training.

    Launch date Downloadable