Abstract:Feature learning is an intensively studied research topic in image classification. Although existing methods like sparse coding, locality-constrained linear coding, fisher vector encoding, etc., have shown their effectiveness in image representation, most of them overlook a phenomenon called thesmall sample size problem, where the number of training samples is relatively smaller than the dimensionality of the features, which may limit the predictive power of the classifier. Subspace learning is a strategy to mitigate this problem by reducing the dimensionality of the features. However, most conventional subspace learning methods attempt to learn a global subspace to discriminate all the classes, which proves to be difficult and ineffective in multi-class classification task. To this end, we propose to learn a local subspace for each sample instead of learning a global subspace for all samples. Our key observation is that, in multi-class image classification, the label of each testing sample is only confused by a few classes which have very similar visual appearance to it. Thus, in this work, we propose a coarse-to-fine strategy, which first picks out such classes, and then conducts a local subspace learning to discriminate them. As the subspace learning method is regularized and conducted within some selected classes, we term it selective regularized subspace learning (SRSL), and we term our classification pipeline selective regularized subspace learning based multi-class image classification (SRSL_MIC). Experimental results on four representative datasets (Caltech-101, Indoor-67, ORL Faces and AR Faces) demonstrate the effectiveness of the proposed method.

Learning an Orthogonal and Smooth Subspace for Image Classification.

Image retrieval based on incremental subspace learning

Orthogonal and Smooth Subspace Based on Sparse Coding for Image Classification.

Incremental semi-supervised subspace learning for image retrieval.

Learning a Locality Preserving Subspace for Visual Recognition.

Learnable Subspace Orthogonal Projection for Semi-supervised Image Classification

Learning a Spatially Smooth Subspace for Face Recognition

Image classification by multimodal subspace learning

Learning subspace kernels for classification

Image Classification by Selective Regularized Subspace Learning

A Semi-supervised Orthogonal Subspace Learning with Spline Embedding

Spatially smooth subspace face recognition using LOG and DOG penalties

Sparse non-negative transition subspace learning for image classification

Subspace learning via Locally Constrained A-optimal nonnegative projection.

A New Robust Subspace Modeling Method

Subspace Learning in The Presence of Sparse Structured Outliers and Noise

Robust Structured Subspace Learning for Data Representation

Exploiting Unsupervised and Supervised Constraints for Subspace Clustering

Low-rank embedded orthogonal subspace learning for zero-shot classification

Subspace-based Multitask Learning Framework for Hyperspectral Imagery Classification

Self-adaptive Subspace Representation from a Geometric Intuition