Fast and Accurate Facial Expression Image Classification and Regression Method Based on Knowledge Distillation

Article Computer Science, Artificial Intelligence

EmoNet: A Transfer Learning Framework for Multi-Corpus Speech Emotion Recognition

Maurice Gerczuk et al.

Summary: This manuscript discusses the topic of multi-corpus Speech Emotion Recognition (SER) from a deep transfer learning perspective. A large corpus called EMOSET is created by assembling emotional speech data from various existing SER corpora. A novel framework called EMONET is developed using a combination of deep ResNet architecture and residual adapters for multi-corpus SER on EMOSET. The introduced residual adapter approach enables efficient training of a multi-domain SER model and leads to improved performance for the majority of the corpora in EMOSET.

IEEE TRANSACTIONS ON AFFECTIVE COMPUTING (2023)

添加到收藏夹

Article Engineering, Multidisciplinary

Distract Your Attention: Multi-Head Cross Attention Network for Facial Expression Recognition

Zhengyao Wen et al.

Summary: This paper presents a novel facial expression recognition network, called Distract your Attention Network (DAN). The proposed method addresses the challenges of inherent similarity and subtle differences among facial expression classes, as well as the need for holistic recognition by encoding high-order interactions among local features. DAN consists of three key components: Feature Clustering Network (FCN), Multi-head Attention Network (MAN), and Attention Fusion Network (AFN). Extensive experiments on three public datasets demonstrate the state-of-the-art performance of DAN in facial expression recognition. The DAN code is publicly available.

BIOMIMETICS (2023)

添加到收藏夹

Article Computer Science, Artificial Intelligence

Multi-level knowledge distillation for low-resolution object detection and facial expression recognition

Tingsong Ma et al.

Summary: This paper proposes a knowledge distillation approach to transfer high-resolution features from a teacher network to a simpler structured student network trained on low-resolution inputs. Experimental results show that the proposed approach achieves higher accuracy than existing models in object detection and facial expression recognition tasks.

KNOWLEDGE-BASED SYSTEMS (2022)

添加到收藏夹

Article Chemistry, Multidisciplinary

Emotion Detection Using Facial Expression Involving Occlusions and Tilt

Awais Salman Qazi et al.

Summary: Facial emotion recognition is an important research topic in pattern recognition, with applications in various fields. This paper proposes a new architecture of a convolutional neural network for the FER system, which achieves higher accuracy through feature map enhancement. An application is developed to identify basic human expressions.

APPLIED SCIENCES-BASEL (2022)

添加到收藏夹

Article Computer Science, Artificial Intelligence

Classifying Emotions and Engagement in Online Learning Based on a Single Facial Expression Recognition Neural Network

Andrey V. V. Savchenko et al.

Summary: This article analyzes the behavior of students in the e-learning environment and proposes a novel pipeline based on video facial processing. By applying face detection, tracking, and clustering techniques, sequences of faces of each student are extracted, and an efficient neural network is used to extract emotional features. The resulting facial features enable fast simultaneous prediction of students' engagement levels, individual emotions, and group-level affect. Additionally, the possibility of creating a summary of a lesson by saving short clips of different emotions and engagement is demonstrated.

IEEE TRANSACTIONS ON AFFECTIVE COMPUTING (2022)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

Face2Exp: Combating Data Biases for Facial Expression Recognition

Dan Zeng et al.

Summary: This study utilizes unlabeled face recognition datasets to enhance facial expression recognition, proposing the Meta-Face2Exp framework to address data bias by learning prior expression knowledge and fitting pseudo labels with a base network and an adaptation network. The circuit feedback mechanism is used to improve the base network and successfully eliminate data bias.

2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022) (2022)

添加到收藏夹

Proceedings Paper Computer Science, Artificial Intelligence

Facial Expression Recognition in the Wild via Deep Attentive Center Loss

Amir Hossein Farzaneh et al.

Summary: The proposed Deep Attentive Center Loss (DACL) method adaptsively selects a subset of significant feature elements to enhance discrimination. By integrating an attention mechanism into CNN's intermediate spatial feature maps, DACL estimates attention weights correlated with feature importance.

2021 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION WACV 2021 (2021)

添加到收藏夹

Article Computer Science, Artificial Intelligence