Convolutional networks and applications in vision

Abstract

Abstract-Intelligent tasks, such as visual perception, auditory perception, and language understanding require the construction of good internal representations of the world (or "features"), which must be invariant to irrelevant variations of the input while, preserving relevant information. A major question for Machine Learning is how to learn such good features automatically. Convolutional Networks (ConvNets) are a biologicallyinspired trainable architecture that can learn invariant features. Each stage in a ConvNets is composed of a filter bank, some non-linearities, and feature pooling layers. With multiple stages, a ConvNet can learn multi-level hierarchies of features. While ConvNets have been successfully deployed in many commercial applications from OCR to video surveillance, they require large amounts of labeled training samples. We describe new unsupervised learning algorithms, and new non-linear stages that allow ConvNets to be trained with very few labeled samples. Applications to visual object recognition and vision navigation for off-road mobile robots are described. I. LEARNING INTERNAL REPRESENTATIONSOne of the key questions of Vision Science (natural and artificial) is how to produce good internal representations of the visual world. What sort of internal representation would allow an artificial vision system to detect and classify objects into categories, independently of pose, scale, illumination, conformation, and clutter? More interestingly, how could an artificial vision system learn appropriate internal representations automatically, the way animals and human seem to learn by simply looking at the world? In the time-honored approach to computer vision (and to pattern recognition in general), the question is avoided: internal representations are produced by a hand-crafted feature extractor, whose output is fed to a trainable classifier. While the issue of learning features has been a topic of interest for many years, considerable progress has been achieved in the last few years with the development of so-called deep learning methods.Good internal representations are hierarchical. In vision, pixels are assembled into edglets, edglets into motifs, motifs into parts, parts into objects, and objects into scenes. This suggests that recognition architectures for vision (and for other modalities such as audio and natural language) should have multiple trainable stages stacked on top of each other, one for each level in the feature hierarchy. This raises two new questions: what to put in each stage? and how to train such deep, multi-stage architectures? Convolutional Networks (ConvNets) are an answer to the first question. Until recently, the answer to the second question was to use gradient-based supervised learning, but recent research in deep learning has produced a number of unsupervised methods which greatly reduce the need for labeled samples. Convolutional NetworksConvolutional Networks [1], [2] are trainable multistage architectures composed of multiple stages. The input and output of each ...

Rotation, Scaling and Deformation Invariant Scattering for Texture Discrimination

“…A logarithmic non-linearity is first applied to invariant scattering coefficients to linearize their power law behavior across scales. This is similar to the normalization strategies used by bag of words [10] and deep neural networks [5].…”

“…Most image representations build localized invariants over small image patches, for example with SIFT descriptors [15]. These invariant coefficients are then aggregated into more invariant global image descriptors, for example with bag of words [10] or multiple layers of deep neural network [4,5]. We follow a similar strategy by first computing invariants over image patches and then aggregating them at the global image scale.…”

“…This is done by applying a logarithm to each coefficient of Sx, which nearly linearizes the dependency upon j 1 and j 2 . This logarithm plays a role which is similar to renormalizations used in bag of words [10] and deep convolution networks [5].…”

“…Hierarchical cascade of invariants [2,3] have been studied to build affine invariant image representations. Deep neural networks provide an architecture to compute such invariants, with a succession of linear filters and "pooling" non-linearities, which are learned from data [4,5].…”

Deep unsupervised network for multimodal perception, representation and classification

“…For instance, it would be interesting to have a global network able to decompose the "concept" of a face as a combination of two eyes, one nose and one mouth, which would be themselves "concepts" from a lower layer. For this reason, using the proposed architecture on top of a standard convolutional network [45,42] may not be an optimal solution, since it would force the network to categorize each image as a whole. This rather suggests to fully integrate the proposed network into a convolutional architecture, which would progressively categorize subparts of an image in a hierarchical way.…”