2019
DOI: 10.1007/978-981-15-1301-5_9
|View full text |Cite
|
Sign up to set email alerts
|

Transfer Learning for Facial Attributes Prediction and Clustering

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
13
0

Year Published

2020
2020
2024
2024

Publication Types

Select...
5
1

Relationship

1
5

Authors

Journals

citations
Cited by 17 publications
(13 citation statements)
references
References 16 publications
0
13
0
Order By: Relevance
“…Among the papers reviewed in the first section of this paper, the highest annotation accuracy on the CelebA data set has been reported by Luca et al [22], where the authors used the DNN architecture MobileNetV2 [19], but without the top classification layers. The architecture of MobileNetV2 contains an initial convolutional layer with 32 filters, followed by 19 residual bottleneck layers.…”
Section: Mobilenetv2mentioning
confidence: 87%
See 4 more Smart Citations
“…Among the papers reviewed in the first section of this paper, the highest annotation accuracy on the CelebA data set has been reported by Luca et al [22], where the authors used the DNN architecture MobileNetV2 [19], but without the top classification layers. The architecture of MobileNetV2 contains an initial convolutional layer with 32 filters, followed by 19 residual bottleneck layers.…”
Section: Mobilenetv2mentioning
confidence: 87%
“…The reason why TNR typically had a higher value than TPR in unbalanced classes is that, in the training data set, there were much more negative than positive examples of particular attributes. We can easily conclude that none of these methods, even the state-of-the-art [22] were capable of classifying the data set with appropriate balance; that is, in a way that both TNR and TPR for each class had similarly high values (i.e., at the level of averaged accuracy of the certain classifier). The highest averaged TPR was obtained for NN 40 (47.54), the second highest was DNN 40 (46.87); the difference between those two was 0.67.…”
Section: Discussionmentioning
confidence: 99%
See 3 more Smart Citations