Multi-label image recognition on MS-COCO 2017 (val)
89.2CP (Classification Precision)ML-GCN
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ML-GCNProtocol=Top 3, Correlation matrix type=Re-weighted2019.04 | 89.2 | — | 64.1 | 74.6 | 90.5 | 66.5 | 76.7 | |
| SRNProtocol=Top 32019.04 | 85.2 | — | 58.8 | 67.4 | 87.4 | 62.5 | 72.9 | |
| ML-GCNProtocol=All, Correlation matrix type=Re-weighted2019.04 | 85.1 | 83 | 72 | 78 | 85.8 | 75.4 | 80.3 | |
| ML-GCNProtocol=Top 3, Correlation matrix type=Binary2019.04 | 84.9 | — | 61.3 | 71.2 | 88.8 | 65.2 | 75.2 | |
| Multi-EvidenceProtocol=Top 32019.04 | 84.5 | — | 62.2 | 70.6 | 89.1 | 64.3 | 74.7 | |
| ResNet-101Protocol=Top 32019.04 | 84.1 | — | 59.4 | 69.7 | 89.1 | 62.8 | 73.6 | |
| ResNet-101Protocol=All2019.04 | 81.6 | 77.1 | 65.4 | 71.2 | 82.7 | 69.9 | 75.8 | |
| ML-GCNProtocol=All, Correlation matrix type=Binary2019.04 | 81.1 | 80.3 | 70.1 | 75.2 | 83.8 | 74.2 | 78.7 | |
| Multi-EvidenceProtocol=All2019.04 | 80.3 | 80.2 | 66.7 | 72.8 | 83.9 | 70.8 | 76.8 | |
| RNN-AttentionProtocol=Top 32019.04 | 79.1 | — | 58.7 | 67.4 | 84 | 63 | 72 | |
| ML-ZSLProtocol=Top 32019.04 | 74.1 | — | 64.5 | 69 | — | — | — | |
| Order-Free RNNProtocol=Top 32019.04 | 71.6 | — | 54.8 | 62.1 | 74.2 | 62.2 | 67.7 | |
| CNN-RNNProtocol=Top 32019.04 | 66 | — | 55.6 | 60.4 | 69.2 | 66.4 | 67.8 | |
| CNN-RNNProtocol=All2019.04 | — | 61.2 | — | — | — | — | — | |
| SRNProtocol=All2019.04 | — | 77.3 | — | — | — | — | — |