Unsupervised Learning of Invariant Feature Hierarchies with Applications to Object Recognition
Top Cited Papers
- 1 June 2007
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 10636919,p. 1-8
- https://doi.org/10.1109/cvpr.2007.383157
Abstract
We present an unsupervised method for learning a hierarchy of sparse feature detectors that are invariant to small shifts and distortions. The resulting feature extractor consists of multiple convolution filters, followed by a feature-pooling layer that computes the max of each filter output within adjacent windows, and a point-wise sigmoid non-linearity. A second level of larger and more invariant features is obtained by training the same algorithm on patches of features from the first level. Training a supervised classifier on these features yields 0.64% error on MNIST, and 54% average recognition rate on Caltech 101 with 30 training samples per category. While the resulting architecture is similar to convolutional networks, the layer-wise unsupervised training procedure alleviates the over-parameterization problems that plague purely supervised learning procedures, and yields good performance with very few labeled training samples.Keywords
This publication has 13 references indexed in Scilit:
- POP: Patchwork of Parts Models for Object RecognitionInternational Journal of Computer Vision, 2007
- Multiclass Object Recognition with Sparse, Localized FeaturesPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2006
- Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene CategoriesPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2006
- Object Recognition with Features Inspired by Visual CortexPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2005
- Distinctive Image Features from Scale-Invariant KeypointsInternational Journal of Computer Vision, 2004
- Semi-Local Affine Parts for Object RecognitionPublished by British Machine Vision Association and Society for Pattern Recognition ,2004
- Sparse coding with an overcomplete basis set: A strategy employed by V1?Published by Elsevier ,2003
- Neocognitron: A new algorithm for pattern recognition tolerant of deformations and shifts in positionPublished by Elsevier ,2003
- Probabilistic visual learning for object detectionPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2002
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998