2019
DOI: 10.1007/s11042-019-7530-7
|View full text |Cite
|
Sign up to set email alerts
|

Multi-stream CNN for facial expression recognition in limited training data

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
6
0

Year Published

2021
2021
2023
2023

Publication Types

Select...
6
3

Relationship

0
9

Authors

Journals

citations
Cited by 33 publications
(6 citation statements)
references
References 30 publications
0
6
0
Order By: Relevance
“…However, MIMO-CxT has several advantages compared to the models proposed in the literature. MIMO-CxT achieved an increase in accuracy of 1.49% and 0.71% when compared to the three-input CNN [26] and the multi-stream CNN [27], respectively. Both models [26], [27] have a deeper and more complex architecture than ours.…”
Section: Discussionmentioning
confidence: 99%
See 1 more Smart Citation
“…However, MIMO-CxT has several advantages compared to the models proposed in the literature. MIMO-CxT achieved an increase in accuracy of 1.49% and 0.71% when compared to the three-input CNN [26] and the multi-stream CNN [27], respectively. Both models [26], [27] have a deeper and more complex architecture than ours.…”
Section: Discussionmentioning
confidence: 99%
“…MIMO-CxT achieved an increase in accuracy of 1.49% and 0.71% when compared to the three-input CNN [26] and the multi-stream CNN [27], respectively. Both models [26], [27] have a deeper and more complex architecture than ours. They both use a large number of convolutional filters and several fully connected layers with a large number of units.…”
Section: Discussionmentioning
confidence: 99%
“…The CNN trained on the difference of landmark coordinates of the neutral and expression image captures the muscle movements corresponding to expression‐specific action units (AUs). Aghamaleki and Chenarlogh 51 also proposed a multi‐stream CNN for FER in static images. The designed CNN fuses high‐level feature maps extracted from the LBP encoded and Sobel convolved horizontal and vertical facial images using a custom VGG16 CNN.…”
Section: Related Workmentioning
confidence: 99%
“…The feature vector concatenation is commonly used to merge and integrate multiple channels or branches in several architectures [14], [15]. The operation that combines features extracted from the VGGinspired branch and features extracted from the pre-trained model is defined as the following formula:…”
Section: Feature Map Fusion Modulementioning
confidence: 99%