RECORD
Separable Batch Normalization for Robust Facial Landmark Localization with Cross-protocol Network Training
Computer Vision and Pattern Recognition · Submitted 17 Jan 2021
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim The experimental results obtained on several well-known datasets demonstrate the effectiveness of the proposed method.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
A big, diverse and balanced training data is the key to the success of deep neural network training. However, existing publicly available datasets used in facial landmark localization are usually much smaller than those for other computer vision tasks. A small dataset without diverse and balanced training samples cannot support the training of a deep network effectively. To address the above issues, this paper presents a novel Separable Batch Normalization (SepBN) module with a Cross-protocol Network Training (CNT) strategy for robust facial landmark localization. Different from the standard BN layer that uses all the training data to calculate a single set of parameters, SepBN considers that the samples of a training dataset may belong to different sub-domains. Accordingly, the proposed SepBN module uses multiple sets of parameters, each corresponding to a specific sub-domain. However, the selection of an appropriate branch in the inference stage remains a challenging task because the sub-domain of a test sample is unknown. To mitigate this difficulty, we propose a novel attention mechanism that assigns different weights to each branch for automatic selection in an effective style. As a further innovation, the proposed CNT strategy trains a network using multiple datasets having different facial landmark annotation systems, boosting the performance and enhancing the generalization capacity of the trained network. The experimental results obtained on several well-known datasets demonstrate the effectiveness of the proposed method.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | A big, diverse and balanced training data is the key to the success of deep neural network training. | background 1.00 | 0.06 | 0.03 | asserted 1.00 |
| s1 | However, existing publicly available datasets used in facial landmark localization are usually much smaller than those for other computer vision tasks. | background 0.78 | 0.07 | 0.02 | asserted 0.92 |
| s2 | A small dataset without diverse and balanced training samples cannot support the training of a deep network effectively. | background 0.81 | 0.05 | 0.03 | asserted 0.98 |
| s3 | To address the above issues, this paper presents a novel Separable Batch Normalization (SepBN) module with a Cross-protocol Network Training (CNT) strategy for robust facial landmark localization. | contribution 0.99 | 0.88 | 0.13 | asserted 0.99 |
| s4 | Different from the standard BN layer that uses all the training data to calculate a single set of parameters, SepBN considers that the samples of a training dataset may belong to different sub-domains. | method 0.87 | 0.21 | 0.03 | hedged possibility 0.62 |
| s5 | Accordingly, the proposed SepBN module uses multiple sets of parameters, each corresponding to a specific sub-domain. | method 0.97 | 0.79 | 0.04 | asserted 1.00 |
| s6 | However, the selection of an appropriate branch in the inference stage remains a challenging task because the sub-domain of a test sample is unknown. | gap 0.56 | 0.09 | 0.02 | asserted 0.98 |
| s7 | To mitigate this difficulty, we propose a novel attention mechanism that assigns different weights to each branch for automatic selection in an effective style. | method 0.57 | 0.89 | 0.05 | asserted 1.00 |
| s8 | As a further innovation, the proposed CNT strategy trains a network using multiple datasets having different facial landmark annotation systems, boosting the performance and enhancing the generalization capacity of the trained network. | contribution 0.63 | 0.89 | 0.12 | asserted 0.99 |
| s9 | The experimental results obtained on several well-known datasets demonstrate the effectiveness of the proposed method. | result 0.95 | 0.95 | 0.68 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
A big, diverse and balanced training data is the key to the success of deep neural network training. However, existing publicly available datasets used in facial landmark localization are usually much smaller than those for other computer vision tasks. A small dataset without diverse and balanced training samples cannot support the training of a deep network effectively. To address the above issues, this paper presents a novel Separable Batch Normalization (SepBN) module with a Cross-protocol Network Training (CNT) strategy for robust facial landmark localization. Different from the standard BN layer that uses all the training data to calculate a single set of parameters, SepBN considers that the samples of a training dataset may belong to different sub-domains. Accordingly, the proposed SepBN module uses multiple sets of parameters, each corresponding to a specific sub-domain. However, the selection of an appropriate branch in the inference stage remains a challenging task because the sub-domain of a test sample is unknown. To mitigate this difficulty, we propose a novel attention mechanism that assigns different weights to each branch for automatic selection in an effective style. As a further innovation, the proposed CNT strategy trains a network using multiple datasets having different facial landmark annotation systems, boosting the performance and enhancing the generalization capacity of the trained network. The experimental results obtained on several well-known datasets demonstrate the effectiveness of the proposed method.
Human review
How the literature uses it
1 arXiv paper cite this in 1 sentence
Papers citing it
Cites it as background 1 paper
-
Jin [ref] introduce cross-protocol network training, where multiple facial landmark datasets are trained simultaneously by sharing a backbone feature encoder and using a different heatmap decoder network for each dataset and thus only shares weights in the CNN feature backbone, but not in the landmark heatmap decoders.
Multi-Domain Multi-Definition Landmark Localization for Small Datasets · Related Works · 0.84
What it cites · 1 sentence without a confident label
Cites as background 6 papers
-
The well-known Batch Normalization method alleviates the convergence problems in deep neural network training effectively [ref] .
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift · Related work · 0.93
-
The rapid development of this research area during the past years produced a variety of effective neural network architecture designs [ref] and loss functions [ref] , which have been instrumental in achieving impressive results.
Adaloss: Adaptive Loss Function for Landmark Localization · Introduction · 0.88
-
The instance normalization method normalizes a feature map along the ($H$, $W$) dimensions [ref] .
Instance Normalization: The Missing Ingredient for Fast Stylization · Related work · 0.86
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.