Research Article
[Retracted] Attention Feature Network Extraction Combined with the Generation Algorithm of Multimedia Image Description
Table 3
Comparison of the scores obtained by different generation algorithms for multimedia image description.
| Algorithm | B@1 | B@4 | METEOR | ROUGEL | CIDEr |
| mRNN | 0.67 | 0.241 | — | — | — | GoogleNIC | — | 0.278 | 0.238 | — | 0.855 | Hard-Attention | 0.718 | 0.251 | 0.231 | — | — | Adaptive | 0.742 | 0.333 | 0.267 | — | 1.085 | DeepVS | 0.625 | 0.231 | 0.196 | — | 0.66 | LSTM + ATTtopdown | 0.8 | 0.367 | 0.269 | 0.569 | 1.158 | SCST : Att2in | — | 0.314 | 0.261 | 0.543 | 1.013 | ATT-FCN | 0.709 | 0.305 | 0.244 | — | — | ERD | — | 0.299 | 0.241 | — | 0.895 | MSM | 0.73 | 0.326 | 0.252 | — | 0.986 | Algorithm put forward in this paper | 0.804 | 0.365 | 0.273 | 0.569 | 1.167 |
|
|