| Essay topic | Author | years | Key words | link |
|---|---|---|---|---|
| Fine-grained facial expression recognition based on multi-scale hierarchical bilinear pooling network | Su Zhiming, Wang Lie, Lan Zhengjie | 2021 | Convolutional neural network; fine-grained expression recognition; multi-scale; hierarchical bilinear pooling; multi-layer feature fusion | https://kns.cnki.net/kcms/detail/31.1289.tp.20210113.1259.002.html |
Summary
Because the subtle inter-class differences and significant intra-class changes of facial expressions make facial expression recognition difficult and result in low recognition rate, a model based on multi-scale bilinear pooling neural network is proposed. First, the global features of facial expressions are extracted through a carefully designed network of three different thickness scales . Then, a hierarchical bilinear pooling layer is introduced to integrate multiple multi-scale cross-layer bilinear features of the same network and different networks to capture part of the feature relationship between different levels , thereby enhancing the model’s representation of subtle facial expressions Discrimination ability. Finally, through layer-by-layer deconvolution and fusion of multi-layer feature information, the problem of missing some key features when the neural network extracts features through multi-layer convolutional layers and pooling layers is solved . The model has the highest recognition rate of 73.725% and 98.28% on the FER2013 and CK+ public data sets. The experimental results show that the proposed method can effectively improve the expression recognition rate and is better than the newer facial expression recognition algorithms such as SPLM, CL, and JNS.
Innovation
The method in this paper has achieved the best results in both CK+ and FER2013 data sets. This is because the method in this article integrates many