Table 1.
The image captioning results of our method and others on the MSCOCO Karpathy test split with cross-entropy loss.
| Method | BLEU-1 | BLEU-2 | BLEU-3 | BLEU-4 | METEOR | ROUGE-L | CIDEr-D | SPICE |
|---|---|---|---|---|---|---|---|---|
| LSTM [20] | - | - | - | 29.6 | 25.2 | 52.6 | 94.0 | - |
| SCST [30] | - | - | - | 30.0 | 25.9 | 53.4 | 99.4 | - |
| Adaptive-Attention [27] | 73.4 | 56.6 | 41.8 | 30.4 | 25.7 | - | 102.9 | - |
| RFNet [40] | 76.4 | 60.4 | 46.6 | 35.8 | 27.4 | 56.5 | 112.5 | 20.5 |
| UpDown [22] | 77.2 | - | - | 36.2 | 27.0 | 56.4 | 113.5 | 20.3 |
| Att2in+RD [43] | - | - | - | 34.3 | 26.4 | 55.2 | 106.1 | 19.7 |
| UpDown+STAM [41] | 77.4 | 61.5 | 47.6 | 36.5 | 27.4 | 56.8 | 114.4 | 20.5 |
| Ours: PW | 77.4 | 61.5 | 47.7 | 36.8 | 28.1 | 57.3 | 117.0 | 21.2 |
| Ours: CW | 77.2 | 61.5 | 47.8 | 36.9 | 28.0 | 57.2 | 117.4 | 21.1 |