Image Annotation on COCO 1000 (test)
69.9R@1semantic-enhanced image and sentence matching model
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| semantic-enhanced image and sentence matching modelBackbone=ResNet-1522017.12 | 69.9 | 92.9 | 97.5 | — | 83.2 | |
| semantic-enhanced image and sentence matching modelBackbone=VGGNet-192017.12 | 66.6 | 91.8 | 96.6 | — | 81.8 | |
| VSE++Backbone=ResNet-1522017.12 | 64.6 | 89.1 | 95.7 | — | 79.4 | |
| VSE++Backbone=VGGNet-192017.12 | 57.2 | 85.1 | 93.3 | — | 74.6 | |
| RRFBackbone=ResNet-1522017.12 | 56.4 | 85.3 | 91.5 | — | 73.9 | |
| 2WayNetBackbone=VGGNet-192017.12 | 55.8 | 75.2 | — | — | — | |
| sm-LSTMBackbone=VGGNet-192017.12 | 53.2 | 83.1 | 91.5 | — | 72 | |
| VQABackbone=VGGNet-192017.12 | 50.5 | 80.1 | 89.7 | — | 68.5 | |
| DSPEBackbone=VGGNet-192017.12 | 50.1 | 79.7 | 89.2 | — | 70.1 | |
| Order-embeddingsTraining Data=COCO2016.09 | 46.7 | — | 88.9 | 2 | — | |
| OEMBackbone=VGGNet-192017.12 | 46.7 | 78.6 | 88.9 | — | 68.6 | |
| M2 (LSTM + SA-FC7)Training Data=COCO2016.09 | 44.6 | 78.4 | 89.5 | 2 | — | |
| MNLMBackbone=VGGNet-192017.12 | 43.4 | 75.7 | 85.8 | — | 63.8 | |
| m-CNNBackbone=VGGNet-192017.12 | 42.8 | 73.1 | 84.1 | — | 64 | |
| m-RNNBackbone=VGGNet-192017.12 | 41 | 73 | 83.5 | — | 57.6 | |
| RNN+FVBackbone=VGGNet-192017.12 | 40.8 | 71.9 | 83.2 | — | 61.8 | |
| FVBackbone=VGGNet-192017.12 | 39.4 | 67.9 | 80.9 | — | 58.3 | |
| DVSABackbone=VGGNet-192017.12 | 38.4 | 69.9 | 80.5 | — | 58.5 |