BLIP-2 ViT-L (zero-shot, 1K test set)

Reported on 6 benchmarks across 2 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision3 results

Image RetrievalonFlickr30k
Recall@1· 2023-01-30
88.6
best: 89.7 (BLIP-2 ViT-G (zero-shot, 1K test set))
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models arXiv:2301.12597
Image RetrievalonFlickr30k
Recall@10· 2023-01-30
98.9
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models arXiv:2301.12597
Image RetrievalonFlickr30k
Recall@5· 2023-01-30
97.6
best: 98.1 (BLIP-2 ViT-G (zero-shot, 1K test set))
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models arXiv:2301.12597

Natural Language Processing3 results

Image-to-Text RetrievalonFlickr30k
Recall@1· 2023-01-30
96.9
best: 97.9 (InternVL-G-FT (finetuned, w/o ranking))
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models arXiv:2301.12597
Image-to-Text RetrievalonFlickr30k
Recall@10· 2023-01-30
100
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models arXiv:2301.12597
Image-to-Text RetrievalonFlickr30k
Recall@5· 2023-01-30
100
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models arXiv:2301.12597