Discriminative Learning and Recognition of Image Set Classes Using Canonical Correlations

By Tae-Kyun Kim, Josef Kittler, Roberto Cipolla. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2007

617 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/6582f4ec2815d2106957215ca2fa298396dde274

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

In Media Technology

Citations in the corpus

Generalized Multiview Analysis: A discriminative latent space (2012, methodology): “CCA is used to match sets of images by maximizing within-set correlation and minimizing between set correlation, which is again a single view scenario with set membership information [13].”
Joint Feature Selection and Subspace Learning for Cross-Modal Retrieval (2016, background): “Recently, several approaches for establishing inter-modal relationships between data from different modalities generally rely on subspace learning, such as Canonical Correlation Analysis [25], [26], Partial Least Squares (PLS) [27] and Bilinear Model (BLM) [28], [29].”
Deep Reconstruction Models for Image Set Classification (2015, methodology): “Image set classification methods commonly model the appearance variability information within images of a set on a geometric surface such as an image set modeled by a subspace [4], [15], a combination of subspaces [5], [16], a point on the Grassmannian [8] or Lie Group [10] of Riemannian manifold.”
Graph embedding discriminant analysis on Grassmannian manifolds for improved image set matching (2011, background): “In contrast to object recognition approaches based on considering one image at a time, there has been a recent surge of interest in techniques based on explicit image set matching [9, 16, 25, 26].”
Cluster Canonical Correlation Analysis (2014, methodology): “To address this, we propose two solutions viz. mean canonical correlation analysis (mean-CCA) and cluster canonical correlation analysis (cluster-CCA).”
Tensor Canonical Correlation Analysis for Action Classification (2007, methodology): “This work was motivated by our previous success [16], where Canonical Correlation Analysis (CCA) is adopted to measure the similarity of any two image sets for robust object recognition.”
Face recognition based on non-corresponding region matching (2011, methodology): “[9] also embedded discriminant analysis into the CCA model, but they applied their model to face video analysis.”
Building Deep Networks on Grassmann Manifolds (2016, methodology): “…weights are set to 93 × 60 , while those of GrNet-2Block Method PaSC-Con PaSC-Han VGGDeepFace [53] 78.82% 68.24% DeepO2P [33] 68.76% 60.14% SPDNet [29] 80.12% 72.83% DCC [39] 75.83 % 67.04% GDA [20] 71.38% 67.49% GGDA [19] 66.71% 68.41% PML [31] 73 are ﬁxed as 93 × 80 , 40 × 30 respectively.”
Human Age Estimation With Regression on Discriminative Aging Manifold (2008, background): “Since the final age estimation comes down to classification, we want to learn a manifold embedding that is sufficiently discriminative [3], [24], [29].”
Canonical Correlation Analysis of Video Volume Tensors for Action Categorization and Detection (2009, background): “Keywords– Action categorization, gesture recognition, canonical correlation analysis, tensor, action detection, incremental subspace learning, spatiotemporal pattern classification.”
Log-Euclidean Metric Learning on Symmetric Positive Definite Manifold with Application to Image Set Classification (2015, methodology): “Discriminant Canonical Correlations (DCC) (Kim et al., 2007), Grassmann Discriminant Analysis (GDA) (Hamm & Lee, 2008), Affine (Convex) Hull based Image Set Distance (AHISD, CHISD) (Cevikalp & Triggs, 2010), Set-to-Set Distance Metric Learning (SSDML) (Zhu et al.”
Projection Metric Learning on Grassmann Manifold with Application to Video based Face Recognition (2015, methodology): “The DCC algorithm attempts to maximize the canonical correlations of withinclass subspace pairs and minimizes the canonical correlations of between-class subspace pairs by transforming the original linear subspaces to low-dimensional ones.”
A Survey of Multi-View Representation Learning (2016, neutral): “Taking multi-view latent space modeling [11, 15] for example, consensus principle is responsible for achieving agreement in a shared lowdimensional representation and complementarity principle guarantees the integration of the complementary knowledge underlying different views.”
Manifold Discriminant Analysis (2009, methodology): “For the set-based methods (MSM, MMD and DCC), we followed the setting similar to that of [15], [26].”
Bregman Divergence-Based Regularization for Transfer Subspace Learning (2010, methodology): “, PCA [22], FLDA [27], [26], LPP [28], MFA [29], DLA [34], and the semisupervised discriminant analysis (SDA) [8].”
Manifold-Manifold Distance with application to face recognition based on image set (2008, methodology): “Representative methods include [10], [17], [27], [28], following the early works in [3], [8].”
Discriminative Deep Metric Learning for Face Verification in the Wild (2014, methodology): “These methods are Discriminant-analysis of Canonical Correlations (DCC) [17], Manifold-Manifold Distance (MMD) [33], Affine Hull based Image Set Distance (AHISD) [5], Convex Hull based Image Set Distance (CHISD) [5], and Sparse Approximated Nearest Points (SANP) [12].”