优化双线性ResNet34的人脸表情识别

Facial Expression Recognition via Optimized Bilinear ResNet34

为了能够更准确且快速地识别人脸表情,提出了一种优化的基于ResNet34网络的双线性结构(OBSR-Net)来进行人脸表情识别. OBSR-Net采用双线性网络结构作为整体框架,主干网络使用ResNet34网络,通过平移不变的方式对局部成对特征交互进行建模,从而提取更加完整有效的特征,同时采用迁移学习的策略来降低人脸表情小样本图像数据集对深度学习方法的限制.此外,在训练过程中使用一种新的通用优化技术,即梯度集中.该方法通过将梯度向量集中到零均值来直接对梯度进行操作,可以看作是一种具有约束损失函数的投影梯度下降方法. OBSR-Net在Fer2013和CK+两个公开数据集上进行实验,分别取得了77.65%和98.82%的识别准确率.实验结果表明,与其他先进的人脸表情识别方法相比, OBSR-Net表现出较强的竞争力.

An optimized bilinear structure based on ResNet34,termed OBSR-Net,is proposed for more accurate and quick facial expression recognition.OBSR-Net adopts a bilinear network structure as its overall framework and incorporates ResNet34 as the backbone network to model the local paired feature interaction by translation invariance,to extract more complete and effective features.At the same time,transfer learning mitigates the limitations imposed by small sample image data sets of facial expressions on deep learning.In addition,gradient concentration,a new general optimization technique,is utilized during the training process.This technique operates directly on gradients by concentrating gradient vectors to zero mean,which can be regarded as a projected gradient descent method with a constrained loss function.Experiments on two public datasets,namely Fer2013 and CK+,reveal that OBSR-Net achieves recognition accuracy of 77.65%and 98.82%,respectively.The experimental results show that OBSR-Net is more competitive than other advanced facial expression recognition methods.