CCMB: A Large-scale Chinese Cross-modal Benchmark

Benchmark Model Rank Results
image-retrieval-on-coco-cnR2D2 (ViT-L/14)#3R@1: 79.1R@10: 98.9R@5: 96.5
image-retrieval-on-coco-cnR2D2 (ViT-B)#6R@1: 75.1R@10: 98.1R@5: 94.2
image-retrieval-on-flickr30k-cnR2D2 (ViT-L/14)#3R@1: 84.4R@10: 98.4R@5: 96.7
image-retrieval-on-flickr30k-cnR2D2 (ViT-B)#8R@1: 78.3R@10: 97.0R@5: 94.6
image-retrieval-on-muge-retrievalR2D2 (ViT-L/14)#4Mean Recall: 77.5R@1: 60.1R@10: 89.4R@5: 82.9
image-retrieval-on-muge-retrievalR2D2 (ViT-B)#8Mean Recall: 68.7R@1: 47.4R@10: 83.5R@5: 75.1