2019年-The implication of spatial temporal changes on facial micro-expression analysis || 论文传送门
论文出发点
当前微表情识别的SOTA方法准确率都还不如人意,很难落地实际应用。这篇论文提供一种思路,让研究者考虑不同数据集的不同参数设置。由于不同微表情数据集的规格通常不一致,体现在如resolution, frame rates, 这篇论文旨在探讨微表情识别中rame rate, resolution及 feature descriptors对于识别效果的影响。
The accuracy of many state-of-the-art methods is still too low to be deployed effectively in a real-world environment. We provide important insights for researchers in this field to consider the settings when conducting new experiment in the future
Due to the inconsistency of facial micro-expression dataset specifications, such as different resolutions and frame rates, we propose to experiment the effects of these variations in micro-expression recognition
数据集(dataset)选择
- CASME II
- SMIC
选择的原因:它们含有很高的帧率,可以很直观地用于测试不同帧率,同时包含了大量的微表情样本和高密度的脸部变化
It has high frame rate which makes it intuitive to test different frame rates. It also contains a high number of micro-expression samples and higher intensity for the facial movements.
原始帧率(frame rate)
CASME II :200 fps
SMIC:100 fps
原始分辨率(resolution)
CASME II:640 x 480
SMIC:190 x 230
特征类型(feature types)
- Local Binary Patterns in Three Orthogonal Planes(LBP-TOP):texture-based features
- 3D Histograms of Oriented Gradient(3DHOG):gradient-based features
- Histogram of Oriented Optical Flow(HOOF): optical flow-based features
分类(classification)
使用SMO(Sequential Minimal Optimization)算法训练SVM。
SMO is able to break down large quadratic programming problems into a series of the smallest possible problems, which are solved analytically and avoids using a time-consuming numerical quadratic programming optimisation as an inner loop. SMO is also able to handle large training sets and is one of the computationally fastest methods of evaluating linear
SVMs.
测评(evaluation)方法
使用F1-Score(因为数据不均衡)
两种验证方法:10-fold交叉验证; Leave-one-subject-out(LOSO)验证
Using the conventional accuracy measure may result in a bias towards classes with large number of samples, hence overestimating the capability of the method. F-Measure micro-average across the whole dataset and is computed based on the total true positives, false negatives and false positives, across 10-fold cross validation and/or Leave-one-subject-out (LOSO) folds
在CASME II上,对于两种验证方法:
- 10-fold 交叉验证: LBP-TOP,200fps,100% of the original resolution,F1-Score 0.637
- Leave-one-subject-out(LOSO)验证:HOOF,200fps,50% resolution,0.439
在SMIC上,对于两种验证方法:
- 10-fold 交叉验证: 3DHOG,50fps,75% of the original resolution,F1-Score 0.624
- Leave-one-subject-out(LOSO)验证:HOOF,100fps,75% resolution,0.614