Wukong: A 100 Million Large-scale Chinese Cross-modal Pre-training Benchmark

Open paper
Benchmark Model Rank Results
image-retrieval-on-coco-cnWukong (ViT-L/14)#7R@1: 74.0R@10: 98.1R@5: 94.4
image-retrieval-on-coco-cnWukong (ViT-B/32)#8R@1: 67.0R@10: 96.7R@5: 91.4
image-retrieval-on-flickr30k-cnWukong (ViT-L/14)#9R@1: 77.4R@10: 97.0R@5: 94.5
image-retrieval-on-flickr30k-cnWukong (ViT-B/32)#10R@1: 67.6R@10: 94.2R@5: 89.6
image-retrieval-on-muge-retrievalWukong (ViT-L/14)#6Mean Recall: 72.1R@1: 52.7R@10: 85.6R@5: 77.9
image-retrieval-on-muge-retrievalWukong (ViT-B/32)#9Mean Recall: 61.2R@1: 39.2R@10: 77.4R@5: 66.9

Rank counts only results with a code link.