FaceNet2ExpNet: Regularizing a Deep Face Recognition Net for Expression Recognition

By Hui Ding, Shaohua Kevin Zhou, Ramalingam Chellappa. IEEE International Conference on Automatic Face & Gesture Recognition, 2016

386 citations recorded by the source corpus.

Relatively small data sets available for expression recognition research make the training of deep networks very challenging. Although fine-tuning can partially alleviate the issue, the performance is still below acceptable levels as the deep features probably contain redundant information from the pretrained domain. In this paper, we present FaceNet2ExpNet, a novel idea to train an expression recognition network based on static images. We first propose a new distribution function to model the high-level neurons of the expression network. Based on this, a two-stage training algorithm is carefully designed. In the pre-training stage, we train the convolutional layers of the expression net, regularized by the face net; In the refining stage, we append fully-connected layers to the pre-trained convolutional layers and train the whole network jointly. Visualization results show that the model trained with our method captures improved high-level expression semantics. Evaluations on four public expression databases, CK+, Oulu- CASIA, TFD, and SFEW demonstrate that our method achieves better results than state-of-the-art.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/1178606d83cc32ca9e99a9ed2aa1b9dd35c11419

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Weihong Deng
2. Xiaojiang Peng
3. Yu Qiao
4. Shan Li
5. Kai-Hong Wang
6. Jianfei Yang
7. Debin Meng
8. Luc Van Gool

In Signal Processing

1. Weihong Deng
2. Shan Li
3. Xiaojiang Peng
4. Jianfei Yang
5. Kai-Hong Wang
6. Lijun Yin
7. Yu Qiao
8. Umur Aybars Ciftci

Citations in the corpus

• Facial Expression Recognition by De-expression Residue Learning (2018, methodology): “Note that only FN2EN [6], PPDN [35] and our DeRL method use the static image for facial expression recognition, while others exploited the temporal information of video sequences.”
• A Novel Space-Time Representation on the Positive Semidefinite Cone for Facial Expression Recognition (2017, background): “Note that the FaceNet2ExpNet [13] is a pure static approach and is reported here as the state-of-the-art of static AFER.”
• Region Attention Networks for Pose and Occlusion Robust Facial Expression Recognition (2019, methodology): “To avoid overfitting on small facial expression datasets, many recent studies [45], [2], [56], [33], [72], [20] utilize face recognition datasets to pre-train a network, and then finetune it on target expression datasets.”
• Deep Facial Expression Recognition: A Survey (2018, methodology): “To eliminate this effect, a two-stage training algorithm FaceNet2ExpNet [68] was proposed.”