Image Retrieval On Crepe Vision Language

評価指標

Recall@1 (HN-Atom + HN-Comp, SC)

Recall@1 (HN-Atom + HN-Comp, UC)

Recall@1 (HN-Atom, UC)

Recall@1 (HN-Comp, UC)

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

モデル名	Recall@1 (HN-Atom + HN-Comp, SC)	Recall@1 (HN-Atom + HN-Comp, UC)	Recall@1 (HN-Atom, UC)	Recall@1 (HN-Comp, UC)	Paper Title	Repository
ViT-B-16 (LAION400M)	37.01	30.81	44.93	59.00	CREPE: Can Vision-Language Foundation Models Reason Compositionally?
Swin-T (CLIP, CC-12M)	-	-	37.3	44.1	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
RN50 (CC12M)	23.26	19.96	34.88	45.27	CREPE: Can Vision-Language Foundation Models Reason Compositionally?
ViT-L-14 (LAION400M)	39.44	33.81	47.86	60.78	CREPE: Can Vision-Language Foundation Models Reason Compositionally?
RN-50 (CLIP, CC-12M)	-	-	36.7	42.9	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
MosaiCLIP (CC-FT)	-	-	40.9	72.4	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
NegCLIP (YFCC-FT)	-	-	39.0	38.8	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
ViT-B-32 (LAION400M)	34.28	28.00	42.75	54.80	CREPE: Can Vision-Language Foundation Models Reason Compositionally?
CLIP-FT (YFCC-FT)	-	-	38.3	36.4	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
RN101 (YFCC15M)	22.74	20.50	39.50	39.56	CREPE: Can Vision-Language Foundation Models Reason Compositionally?
ViT-B-16+240 (LAION400M)	37.32	32.26	46.53	60.19	CREPE: Can Vision-Language Foundation Models Reason Compositionally?
CLIP-FT (CC-FT)	-	-	35.6	45.8	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
CLIP (YFCC-FT)	-	-	39.5	39.8	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
CLIP (CC-FT)	-	-	35.0	45.1	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
Random	9.09	9.09	20.00	14.29	CREPE: Can Vision-Language Foundation Models Reason Compositionally?
RN-50 (NegCLIP, CC-12M)	-	-	41.4	82.0	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
RN-50 (MosaiCLIP, CC-12M)	-	-	44.4	92.6	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
NegCLIP (CC-FT)	-	-	37.5	53.1	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
Swin-T (MosaiCLIP, CC-12M)	-	-	44.5	92.1	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-
Swin-T (NegCLIP, CC-12M)	-	-	39.6	80.3	Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality	-

0 of 22 row(s) selected.