Face recognition with learning-based descriptor

By Zhimin Cao, Qi Yin, Xiaoou Tang. 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010

481 citations recorded by the source corpus.

We present a novel approach to address the representation issue and the matching issue in face recognition (verification). Firstly, our approach encodes the micro-structures of the face by a new learning-based encoding method. Unlike many previous manually designed encoding methods (e.g., LBP or SIFT), we use unsupervised learning techniques to learn an encoder from the training examples, which can automatically achieve very good tradeoff between discriminative power and invariance. Then we apply PCA to get a compact face descriptor. We find that a simple normalization mechanism after PCA can further improve the discriminative ability of the descriptor. The resulting face representation, learning-based (LE) descriptor, is compact, highly discriminative, and easy-to-extract. To handle the large pose variation in real-life scenarios, we propose a pose-adaptive matching method that uses pose-specific classifiers to deal with different pose combinations (e.g., frontal v.s. frontal, frontal v.s. left) of the matching face pair. Our approach is comparable with the state-of-the-art methods on the Labeled Face in Wild (LFW) benchmark (we achieved 84.45% recognition rate), while maintaining excellent compactness, simplicity, and generalization ability across different datasets.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/6043006467fb3fd1e9783928d8040ee1f1db1f3a

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Learning Discriminant Face Descriptor (2014, methodology): “Most learning-based descriptors [28], [30], [31] focus on the improvement at the second or the third step.”
A Comprehensive Study on Center Loss for Deep Face Recognition (2019, methodology): “To improve the face representations, several papers investigated SIFT (Lowe 2004), HoG (Dalal and Triggs 2005), LBP (Ahonen et al. 2006) and its variants (Lu et al. 2015; Duan et al. 2017), and learned features (Cao et al. 2010).”
Probabilistic Elastic Matching for Pose Variant Face Verification (2013, methodology): “Predominant recent works focused on the less restricted protocol [5, 8, 25] and least restricted protocol [18, 33, 3], which have pushed the recognition accuracy to be as high as 0.”
PCCA: A new approach for distance learning from sparse pairwise constraints (2012, background): “The much more elaborate and face-specific LE descriptor [3] gives only 83.”
Knowledge-based Systems (2009, unknown): “For this reason, to tackle the aggregation problem that we pose in this paper, we resort to the tools in the social choice literature.”
Coupled information-theoretic encoding for face photo-sketch recognition (2011, methodology): “We ﬁnd that sampling the gradients I u and I v results in a better descriptor than sampling the intensities [ 5].”
Learning LBP structure by maximizing the conditional mutual information (2015, methodology): “In [31], Cao et al. utilized a random projection tree to encode LBP features, and used PCA to reduce the dimensionality .”
Multi-View Perceptron: a Deep Model for Learning Face Identity and View Representations (2014, unknown): “6 LE [3]+LDA 93.”
A Comprehensive Survey on Pose-Invariant Face Recognition (2015, unknown): “[Cao et al. 2010] Multiple LE+comp Feature Protocol4 84.”
Hybrid Deep Learning for Face Verification (2013, methodology): “Face recognition has been extensively studied in recent decades [29, 28, 30, 1, 16, 5, 33, 12, 6, 3, 7, 25, 34].”
Learning hierarchical representations for face verification with convolutional deep belief networks (2012, methodology): “[3] form a pixel-level feature representation by circular sampling similar to LBP, then quantize these feature vectors using random-projection trees.”
Naive-Deep Face Recognition: Touching the Limit of LFW Benchmark or Not? (2015, background): “The amount of data expanded 100 times from 2010 to 2014 (e.g., from about 10 thousand training samples in Multiple LE [3] to 4 millions images in DeepFace [14]).”
AutoCaption: Automatic caption generation for personal photos (2014, background): “Currently our system runs face detection [26, 20], face recognition [4], proxemics detection [24], scene classification [23] and landmarks recognition [2].”
Learning Deep Face Representation (2014, background): “LTP [12], TPLBP, FPLBP [13], CLBP [14], LGBPHS [15], LE [16]).”
Surpassing Human-Level Face Verification Performance on LFW with GaussianFace (2014, methodology): “Even for features learned from data [10, 24], the algorithm parameters (such as the depth of random projection tree, or the number of centers in k-means) also need to be speciﬁed by users.”
When Face Recognition Meets with Deep Learning: An Evaluation of Convolutional Neural Networks for Face Recognition (2015, methodology): “Thus the features fed into MeL can be hand-crafted (SIFT, LBP, Fisher vector [19]) or learning-based (CNN [24, 31] and the learning-based descriptor [3]).”
A Deep Sum-Product Architecture for Robust Facial Attributes Analysis (2013, background): “As face attributes are becoming more important for face recognition, in the future work, we will explore the advantage to combine face attributes with traditional face recognition approaches [23, 24, 3, 26].”
Face Alignment at 3000 FPS via Regressing Local Binary Features (2014, methodology): “Ensemble trees can be used as a codebook for efficient encoding [22] or learning better descriptors [6] [33].”