Hierarchical Neural Networks for Image Interpretation

By Sven Behnke. 2003

232 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks (2015, methodology): “Beyond simple sequences, RNNs exist in full multi-dimensional variants, such as those introduced by Graves and Schmidhuber [11] for handwriting recognition, or as hierarchical recurrent pyramids (the Neural Abstraction Pyramid [2]).”
Multi-column deep neural networks for image classification (2012, methodology): “Table 3 shows that more preprocessing results in lower MCDNN error.”
Striving for Simplicity: The All Convolutional Net (2014, methodology): “For a ﬁrst step in that direction we study properties of our network using a deconvolutional approach similar to Zeiler & Fergus (2014).”
Fast image scanning with deep max-pooling convolutional neural networks (2013, methodology): “They were first successfully applied to relatively small tasks such as digit recognition [2], image interpretation [3] and object recognition [4].”
Deep learning in neural networks: An overview (2014, methodology): “See also more recent methods using UL to improve subsequent SL (e.g., Behnke, 1999, 2003a; Escalante-B & Wiskott, 2013).”
High-Performance Neural Networks for Visual Object Classification (2011, neutral): “…Serre et al. , 2007; Mutch and Lowe, 2008), whose ﬁlters are ﬁxed, in contrast to those of Convolutional Neural Networks (CNNs) (LeCun et al. , 1998; Behnke, 2003; Simard et al. , 2003), whose weights (ﬁlters) are randomly initialized and changed in a supervised way using back-propagation (BP).”
On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models (2015, methodology): “In particular, many recent computer vision contests were won by fully supervised Max-Pooling Convolutional NNs (MPCNNs), which consist of alternating convolutional [58, 19] and max-pooling [296] layers topped off by standard fully connected output layers.”
One Big Net For Everything (2018, background): “If there are large 2-dimensional inputs such as video images, the LSTM may have a front-end [95] in form of a convolutional neural net (CNN) [11, 33, 97, 4, 45, 52, 8, 76] implemented on fast GPUs [8, 76].”
Self-Delimiting Neural Networks (2012, background): “…Neocognitron [13]; they sometimes (but not always) proﬁt from sparse network connectivity and techniques such as weight sharing & convolution [38, 1], max-pooling [49], and contrast enhancement like the one automatically generated by unsupervised Predictability Minimization [53, 64, 67].”
RGB-D object recognition and pose estimation based on pre-trained convolutional neural network features (2015, methodology): “Deep convolutional neural networks (CNN) [4]–[6] became the dominant method in the ImageNet large scale image classification challenge [7] since the seminal work of Krizhevsky et al. [8].”
Large-scale object recognition with CUDA-accelerated hierarchical neural networks (2009, unknown): “Similar to the Neural Abstraction Pyramid [6], there are L ≥ 1 regular layers and one output layer.”
Steel defect classification with Max-Pooling Convolutional Neural Networks (2012, unknown): “In this work we focus on the two last pipeline stages and propose an approach based on Max-Pooling Convolutional Neural Networks (MPCNN) [1], [2], [3], [4], [5], that learn the features directly from labeled images using supervised learning.”
Discovering hierarchical speech features using convolutional non-negative matrix factorization (2003, methodology): “For images, such hierarchical models have been discovered by unsupervised learning techniques and successfully used for object recognition tasks in the Neural Abstraction Pyramid architecture [8], [9].”
Mitosis Detection in Breast Cancer Histology Images with Deep Neural Networks (2013, unknown): “Introduced in 1980 [6] and gradually improved over the next two decades [11,1,17], they unfold their full potential when combined with MP and made both deep and wide [3,4].”