Supervised translation-invariant sparse coding
Jianchao Yang, Kai Yu, Thomas S. Huang · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · 2010
In this paper, we propose a novel supervised hierarchical sparse coding model based on local image descriptors for classification tasks. The supervised dictionary training is performed via back-projection, by minimizing the training error of classifying the image level features, which are extracted by max pooling over the sparse codes within a spatial pyramid. Such a max pooling procedure across multiple spatial scales offer the model translation invariant properties, similar to the Convolutional Neural Network (CNN). Experiments show that our supervised dictionary improves the performance of the proposed model significantly over the unsupervised dictionary, leading to state-of-the-art performance on diverse image databases. Further more, our supervised model targets learning linear features, implying its great potential in handling large scale datasets in real applications.
Cited by 355 works. doi:10.1109/cvpr.2010.5539958
Authors
References
- Gradient-based learning applied to document recognition
- What is the best multi-stage architecture for object recognition?
- Eigenfaces vs. Fisherfaces: recognition using class specific linear projection
- Convolutional deep belief networks for scalable unsupervised learning of hierarchical representations
- $rm K$-SVD: An Algorithm for Designing Overcomplete Dictionaries for Sparse Representation
- Linear spatial pyramid matching using sparse coding for image classification
- Robust Face Recognition via Sparse Representation
- Efficient sparse coding algorithms
- Self-taught learning
- For most large underdetermined systems of linear equations the minimal 𝓁 1 ‐norm solution is also the sparsest solution