Text-to-Image Retrieval on MS-COCO 1k (test)
64.7R@1CRCL
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CRCLNoise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 64.7 | — | 90.6 | 95.9 | 525.6 | |
| MSCNNoise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 64.3 | — | 90.4 | 95.8 | 524.6 | |
| BiCroNoise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 63.7 | — | 90.3 | 95.7 | 523.2 | |
| VSRN + AOQD=2048 x 22021.01 | 63.5 | 45.6 | 90.5 | 95.8 | — | |
| VSRN + AOQ2021.01 | 63.5 | 45.6 | — | — | — | |
| CRCLNoise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 63.3 | — | 90.3 | 95.7 | 521.5 | |
| VSRND=20482021.01 | 62.8 | 42.4 | 89.7 | 95.1 | — | |
| VSRN2021.01 | 62.8 | 42.4 | — | — | — | |
| BiCroNoise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 61.8 | — | 89.2 | 94.9 | 517.1 | |
| DECLNoise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 61.6 | — | 89.1 | 95.2 | 517.3 | |
| NCRNoise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 61.5 | — | 89.3 | 95.1 | 516.6 | |
| DECLNoise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 61.5 | — | 89.2 | 95.3 | 517.1 | |
| CRCLNoise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 60.8 | — | 89 | 95.1 | 514.2 | |
| NCRNoise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 60.7 | — | 88.5 | 95 | 513.9 | |
| MSCNNoise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 60.3 | — | 88.5 | 94.4 | 510.9 | |
| DECLNoise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 59.6 | — | 87.9 | 94.5 | 508.2 | |
| MSCNNoise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 59.2 | — | 87.1 | 92.8 | 506.5 | |
| BiCroNoise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 58.3 | — | 87.2 | 93.9 | 505.5 | |
| NCRNoise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 57.9 | — | 87 | 94.1 | 503.3 | |
| CRCLNoise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 57.5 | — | 86.8 | 93.7 | 501.8 | |
| VSE∞Noise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 57.4 | — | 86.5 | 92.8 | 500.5 | |
| PVSEK=2, D=1024 x 22021.01 | 55.2 | 43.7 | 86.5 | 93.7 | — | |
| PVSEK=22021.01 | 55.2 | 43.6 | — | — | — | |
| PCME (mu only)D=10242021.01 | 54.6 | 45.9 | 86.3 | 93.8 | — | |
| PCMED=1024 x 22021.01 | 54.6 | 46 | 86.3 | 93.8 | — | |
| PCMEvariant=mu only2021.01 | 54.6 | 45.9 | — | — | — | |
| PCMEvariant=stochastic2021.01 | 54.6 | 46 | — | — | — | |
| PVSEK=4, D=1024 x 42021.01 | 54.1 | 42.7 | 85.5 | 92.9 | — | |
| PVSE + SHMK=1, D=1024 x 12021.01 | 53.6 | 42.4 | 85.5 | 93 | — | |
| PVSEK=1, D=10242021.01 | 53.5 | 41.9 | 85.1 | 92.7 | — | |
| PVSEK=12021.01 | 53.5 | 41.8 | — | — | — | |
| PVSE + SHMK=2, D=1024 x 22021.01 | 53.1 | 39.4 | 85.4 | 93 | — | |
| MSCNNoise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 52.7 | — | 83 | 90.9 | 481.2 | |
| VSE++D=10242021.01 | 52 | — | 84.3 | 92 | — | |
| VSE++2021.01 | 52 | — | — | — | — | |
| DECLNoise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 52 | — | 83 | 91.3 | 480.1 | |
| SGRNoise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 42.5 | — | 77.7 | 88.2 | 434 | |
| SAFNoise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 38.2 | — | 74 | 85.5 | 406.5 | |
| SGRNoise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 37 | — | 72.8 | 85.1 | 406.4 | |
| SGRNoise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 33.8 | — | 68.6 | 81.7 | 381.3 | |
| SAFNoise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 33.7 | — | 69.4 | 82.5 | 383 | |
| VSE∞Noise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 31.4 | — | 63.8 | 75 | 399.9 | |
| SAFNoise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 31.1 | — | 65.6 | 80.5 | 348.7 | |
| VSE∞Noise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 26 | — | 60.1 | 76.3 | 339.7 | |
| SGRNoise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 25.3 | — | 58.2 | 72.6 | 319.1 | |
| SAFNoise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 24.7 | — | 57.1 | 73 | 310.6 | |
| BiCroNoise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 22.6 | — | 53 | 67.2 | 340.1 | |
| SCANNoise Rate=20%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 20.6 | — | 55.6 | 73.5 | 322.6 | |
| VSE∞Noise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 19.2 | — | 50.5 | 68 | 288.8 | |
| SCANNoise Rate=40%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 18.9 | — | 51.1 | 69.9 | 314.4 | |
| SCANNoise Rate=60%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 16.8 | — | 47.8 | 66.4 | 293.4 | |
| NCRNoise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 15.1 | — | 38.1 | 49.8 | 244.8 | |
| SCANNoise Rate=80%, Visual Feature=BUTD, Backbone=Bi-GRU, Evaluation=5-fold average2023.10 | 13.8 | — | 41.1 | 58.6 | 255.1 |