Weakly supervised facial expression recognition via transferred DAL-CNN and active incremental learning

Xu, Ying; Liu, Jian; Zhai, Yikui; Gan, Junying; Zeng, Junying; Cao, He; Scotti, Fabio; Piuri, Vincenzo; Labati, Ruggero Donida

doi:10.1007/s00500-019-04530-1

Cited by 15 publications

(3 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Due to the decline of the functions of the elderly in all aspects of the body, the current social focus is more on the physical health of the elderly who live alone without care. In people's daily life, the posture of the human body is closely related to the health of the human body; the injury caused by the fall of the elderly is very serious and even endangers life safety; serious harm caused may include soft tissue injury, muscle tear, collision, cerebral hemorrhage or accidental head injury, and spinal fracture [6,7].…”

Section: Introductionmentioning

confidence: 99%

Accurate Detection of Intelligent Running Posture Based on Artificial Intelligence Sensor

Zhang¹,

Cheng²

2022

Journal of Sensors

View full text Add to dashboard Cite

In order to solve the problem of low accuracy of human posture recognition during motion, the author proposes a human posture recognition and detection method based on multiple sensors. The method uses acceleration sensor, angular velocity sensor, and single-chip microcomputer to collect data; uses time domain and frequency domain analysis methods to analyze the collected data; and then uses Bayesian classifier to classify and identify the current motion posture of the human body. The obtained results are as follows: in the traditional method, with the increase of subjects, the accuracy of the classification method continues to decline; when the number of subjects reaches 50, the accuracy rate is only 60%; in the experiment, the zero crossing times of the acceleration signal axis are selected, and the axis area of the angular velocity signal is used as the characteristic item to distinguish the two different attitudes of going upstairs and going downstairs; the discrimination can reach about 90%. The authors combine the data collected by the acceleration sensor and the angular velocity sensor to perform feature extraction and classification, and the classification accuracy can reach more than 95%. It is proved that the method proposed by the authors can perform human gesture recognition very quickly and has a high accuracy rate.

show abstract

Section: Introductionmentioning

confidence: 99%

Accurate Detection of Intelligent Running Posture Based on Artificial Intelligence Sensor

Zhang¹,

Cheng²

2022

Journal of Sensors

View full text Add to dashboard Cite

show abstract

“…Heo et al [ 18 ] proposed an occlusion-aware spatial attention transformer (OSAT) architecture based on a visual transformer (ViT), CutMix strengthening, and occlusion mask predictor (OMP) to solve the occlusion problem. Xu et al [ 19 ] proposed a double-active-layer-based CNN to recognize facial expressions with high accuracy by learning robust and discriminative features from data. However, because of the occlusion caused by the hand and the great resemblance of the commodities, the accuracy of goods recognition using conventional methods is low.…”

Section: Related Workmentioning

confidence: 99%

Recognition of Occluded Goods under Prior Inference Based on Generative Adversarial Network

Cao

Xie

Liu

et al. 2023

Sensors

View full text Add to dashboard Cite

Aiming at the recognition of intelligent retail dynamic visual container goods, two problems that lead to low recognition accuracy must be addressed; one is the lack of goods features caused by the occlusion of the hand, and the other is the high similarity of goods. Therefore, this study proposes an approach for occluding goods recognition based on a generative adversarial network combined with prior inference to address the two abovementioned problems. With DarkNet53 as the backbone network, semantic segmentation is used to locate the occluded part in the feature extraction network, and simultaneously, the YOLOX decoupling head is used to obtain the detection frame. Subsequently, a generative adversarial network under prior inference is used to restore and expand the features of the occluded parts, and a multi-scale spatial attention and effective channel attention weighted attention mechanism module is proposed to select fine-grained features of goods. Finally, a metric learning method based on von Mises–Fisher distribution is proposed to increase the class spacing of features to achieve the effect of feature distinction, whilst the distinguished features are utilized to recognize goods at a fine-grained level. The experimental data used in this study were all obtained from the self-made smart retail container dataset, which contains a total of 12 types of goods used for recognition and includes four couples of similar goods. Experimental results reveal that the peak signal-to-noise ratio and structural similarity under improved prior inference are 0.7743 and 0.0183 higher than those of the other models, respectively. Compared with other optimal models, mAP improves the recognition accuracy by 1.2% and the recognition accuracy by 2.82%. This study solves two problems: one is the occlusion caused by hands, and the other is the high similarity of goods, thus meeting the requirements of commodity recognition accuracy in the field of intelligent retail and exhibiting good application prospects.

show abstract

“…Pose-invariant FER has drawn great research attention due to its potential value in practical application, such as non-invasive identification of uncooperative objects, human-computer interaction, digital entertainment, driver fatigue surveillance and others [1][2][3][4]. Over the past decades, many effective and feasible methods have been proposed for FER, where the facial images are regard as relatively static and remain in a frontal or near-frontal state.…”

Section: Introductionmentioning

confidence: 99%

Deep Spatial and Channel Sliding Attention Patches for Pose-invariant Facial Expression Recognition

Liu

Chen

et al. 2023

Preprint

View full text Add to dashboard Cite

Pose-invariant facial expression recognition (FER) is an import yet challenging research topics in computer vision, especially with the introduction of pose change and self-occlusion, which makes the recognition results changing from one observational angle to another. In this paper, a newly deep sliding patch combined with spatial-channel attention network is designed for pose-invariant FER. The new network architecture consists of three main components, namely, the sliding patch (SP) model, deep channel squeeze-and-extraction model (SE) and spatial sliding patch attention (SPA) model. The sliding patch (SP) model is devised to find the best patch size and stride for pose-invariant FER. The sliding patch attention (SPA) model can guide the network pay more attention to the current local patches and adaptively estimate the significance of each local sliding patch. The squeeze-and-extraction (SE) operation was implemented prior to SPA model whose aims to shrink the number of feature maps and provide more salient semantic feature maps for SPA model. Extensive experiments were implemented on four pose robust expression datasets, namely, BU3DFEP1, BU3DFEP2, Multi-PIE, Pose-RAFDB, Pose-Affect, and the experimental results indicated the validity and feasibility of SPA-SE network.

show abstract

Weakly supervised facial expression recognition via transferred DAL-CNN and active incremental learning

Cited by 15 publications

References 29 publications

Accurate Detection of Intelligent Running Posture Based on Artificial Intelligence Sensor

Accurate Detection of Intelligent Running Posture Based on Artificial Intelligence Sensor

Recognition of Occluded Goods under Prior Inference Based on Generative Adversarial Network

Deep Spatial and Channel Sliding Attention Patches for Pose-invariant Facial Expression Recognition

Contact Info

Product

Resources

About