Facial Expression Recognition in the Wild via Deep Attentive Center Loss

Farzaneh, Amir Hossein; Qi, Xiaojun

doi:10.1109/wacv48630.2021.00245

Cited by 197 publications

(81 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wang et al [36] proposed a Region Attention Network (RAN) to capture facial regions for occlusion and pose variant FER. Farzaneh and Qi [11] introduced a Deep Attentive Center Loss (DACL) method to estimate the attention weight for the features for enhancing the discrimination. A sparse center loss was designed to achieve intra-class compactness and inter-class separation with the weighted features.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

POSTER: A Pyramid Cross-Fusion Transformer Network for Facial Expression Recognition

Zheng¹,

Mendieta²,

Chen³

2022

Preprint

View full text Add to dashboard Cite

Facial Expression Recognition (FER) has received increasing interest in the computer vision community. As a challenging task, there are three key issues especially prevalent in FER: inter-class similarity, intra-class discrepancy, and scale sensitivity. Existing methods typically address some of these issues, but do not tackle them all in a unified framework. Therefore, in this paper, we propose a two-stream Pyramid crOss-fuSion TransformER network (POSTER) that aims to holistically solve these issues. Specifically, we design a transformer-based cross-fusion paradigm that enables effective collaboration of facial landmark and direct image features to maximize proper attention to salient facial regions. Furthermore, POSTER employs a pyramid structure to promote scale invariance. Extensive experimental results demonstrate that our POSTER outperforms SOTA methods on RAF-DB with 92.05 %, FERPlus with 91.62 %, AffectNet (7 cls) with 67.31 %, and Affect-Net (8 cls) with 63.34 %, respectively. Code is available at https: //github.com/zczcwh/POSTER

show abstract

Section: Related Workmentioning

confidence: 99%

“…We set the batch size to 100 with a learning rate of 4 × 10 −5 . Unlike many methods [11,19] that rely on complicated loss, we use the standard label smoothing cross-entropy loss. Evaluation on RAF-DB: Table 1 compares POSTER with previous methods on the RAF-DB dataset.…”

Section: Implementation Detailsmentioning

confidence: 99%

POSTER: A Pyramid Cross-Fusion Transformer Network for Facial Expression Recognition

Zheng¹,

Mendieta²,

Chen³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…At the testing time, the central crops are fed into a trained model on AffectNet-7 and DAF_DB for 60 epochs; the initial learning rate is 0.01 with a reduction in a factor of 10 every 20 epochs; the batch size is set to 32 for AffectNet-7 and 16 for DAF_DB, and in experiments, the model is trained with Pytorch on NVIDIA GeForce GTX 1650 GPU with 16 GB RAM. Notably, the pre-training strategy is adapted for saving the total training time and can obtain superior performance [ 51 ]. In this paper, the proposed model will be pre-trained on a face dataset MS-CELEB-1M [ 52 ], and then fine-tuned on FER datasets, owing to the similarity between the domain of FER and the face recognition (FR) task.…”

Section: Resultsmentioning

confidence: 99%

Facial Expression Recognition: One Attention-Modulated Contextual Spatial Information Network

Zhu

Zhou

2022

Entropy

View full text Add to dashboard Cite

Facial expression recognition (FER) in the wild is a challenging task due to some uncontrolled factors such as occlusion, illumination, and pose variation. The current methods perform well in controlled conditions. However, there are still two issues with the in-the-wild FER task: (i) insufficient descriptions of long-range dependency of expression features in the facial information space and (ii) not finely refining subtle inter-classes distinction from multiple expressions in the wild. To overcome the above issues, an end-to-end model for FER, named attention-modulated contextual spatial information network (ACSI-Net), is presented in this paper, with the manner of embedding coordinate attention (CA) modules into a contextual convolutional residual network (CoResNet). Firstly, CoResNet is constituted by arranging contextual convolution (CoConv) blocks of different levels to integrate facial expression features with long-range dependency, which generates a holistic representation of spatial information on facial expression. Then, the CA modules are inserted into different stages of CoResNet, at each of which the subtle information about facial expression acquired from CoConv blocks is first modulated by the corresponding CA module across channels and spatial locations and then flows into the next layer. Finally, to highlight facial regions related to expression, a CA module located at the end of the whole network, which produces attentional masks to multiply by input feature maps, is utilized to focus on salient regions. Different from other models, the ACSI-Net is capable of exploring intrinsic dependencies between features and yielding a discriminative representation for facial expression classification. Extensive experimental results on AffectNet and RAF_DB datasets demonstrate its effectiveness and competitiveness compared to other FER methods.

show abstract

“…Accuracy, % 8 classes 7 classes PSR (VGG-16) [20] 60.68 -ARM (ResNet-18) [21] 59.75 64.49 RAN [22] 59.5 -PAENet [8] -65.29 DACL [23] -65.20 EmotionNet (InceptionResNet-v1) [8] -64.74 CPG [24] - dataset. Moreover, we report the results of models after the first training stage, in which only the weights of the classification head were learned, while the other part of the model remains the same as in the pre-trained face recognition CNN (left part of Fig.…”

Section: Methodsmentioning

confidence: 99%

Facial expression and attributes recognition based on multi-task learning of lightweight neural networks

Savchenko

2021

Preprint

View full text Add to dashboard Cite

In this paper, we examine the multi-task training of lightweight convolutional neural networks for face identification and classification of facial attributes (age, gender, ethnicity) trained on cropped faces without margins. It is shown that it is still necessary to fine-tune these networks in order to predict facial expressions. Several models are presented based on MobileNet, EfficientNet and RexNet architectures. It was experimentally demonstrated that our models are characterized by the state-of-the-art emotion classification accuracy on AffectNet dataset and near state-ofthe-art results in age, gender and race recognition for UTKFace dataset. Moreover, it is shown that the usage of our neural network as a feature extractor of facial regions in video frames and concatenation of several statistical functions (mean, max, etc.) leads to 4.5% higher accuracy than the previously known state-of-the-art single models for AFEW and VGAF datasets from the EmotiW challenges. The models and source code are publicly available at https://github.com/HSE-asavchenko/ face-emotion-recognition.

show abstract

Facial Expression Recognition in the Wild via Deep Attentive Center Loss

Cited by 197 publications

References 26 publications

POSTER: A Pyramid Cross-Fusion Transformer Network for Facial Expression Recognition

POSTER: A Pyramid Cross-Fusion Transformer Network for Facial Expression Recognition

Facial Expression Recognition: One Attention-Modulated Contextual Spatial Information Network

Facial expression and attributes recognition based on multi-task learning of lightweight neural networks

Contact Info

Product

Resources

About