Noisy Correspondence Learning with Self-Reinforcing Errors Mitigation

Benchmark Model Rank Results
cross-modal-retrieval-with-noisy-1SREMR-Sum: 372.2Image-to-text R@1: 40.9Image-to-text R@5: 67.5
cross-modal-retrieval-with-noisy-2SREMR-Sum: 507.8Image-to-text R@1: 79.5Image-to-text R@5: 94.2
cross-modal-retrieval-with-noisy-3SREMR-Sum: 524.1Image-to-text R@1: 78.5Image-to-text R@5: 96.8