基于深度学习的表情动作单元识别综述
|
|
邵志文, 周勇, 谭鑫, 马利庄, 刘兵, 姚睿
|
Survey of Expression Action Unit Recognition Based on Deep Learning
|
|
SHAO Zhi-wen, ZHOU Yong, TAN Xin, MA Li-zhuang, LIU Bing, YAO Rui
|
|
表4 基于深度学习的AU检测代表性方法总结
|
| |
|
| 方法 | 基于已有 模型的 迁移学习 | 基于已有 标签的 迁移学习 | 基于 域映射的 迁移学习 | 特征点 辅助的 区域学习 | 自适应 区域学习 | 像素级 关联学习 | AU级 关联学习 | 时域 关联学习 | BP4D[26]/DISFA[27] |
|---|
| DRML[68] | | | | | √ | | | | 0.483/0.267 | | EAC-Net[61] | √ | | | √ | | | √ | | 0.559/0.485 | | R-T1[65] | √ | | | √ | | | √ | √ | 0.661/0.513 | | DSIN[77] | | | | √ | | | √ | | 0.589/0.536 | | LP-Net[72] | √ | | | √ | | √ | | | 0.610/0.569 | | MLCR[82] | | √ | | | | | √ | | 0.598/— | | SRERL[79] | √ | | | √ | | | √ | | 0.629/0.559 | | ARL[24] | | | | | √ | √ | √ | | 0.611/0.587 | | PAttNet[70] | | | | √ | | | | | 0.626/— | | D-PAttNet[71] | | | | √ | | | | √ | 0.641/— | | TAE[60] | | | √ | | √ | | | | 0.603/0.515 | | OF-Net[91] | | √ | | | | | | √ | 0.597/0.537 | | AU R-CNN[64] | | √ | | √ | | | | | 0.630/0.513 | | AU-GCN[81] | | | | √ | | | √ | | 0.628/0.550 | | JÂA-Net[52] | | √ | | √ | | | √ | | 0.624/0.635 | | UGN-B[84] | √ | | | | | | √ | | 0.633/0.600 | | Transformer[78] | √ | | | √ | | | √ | | 0.642/0.615 | | HMP-PS[85] | √ | | | | | | √ | | 0.634/0.610 |
|
|
|