https://www2.econ.iastate.edu/tesfatsi/DeepLearningInNeuralNetworksOverview.JSchmidhuber2015.pdf

Deep learning in neural networks

In this work we address the task of semantic image segmentation with Deep Learning and make three main contributions that are experimentally shown to have substantial practical merit. First , we highlight convolution with upsampled filters, or ‘atrous convolution’, as a powerful tool in dense prediction tasks. Atrous convolution allows us to explicitly control the resolution at which feature responses are computed within Deep Convolutional Neural Networks. It also allows us to effectively enlarge the field of view of filters to incorporate larger context without increasing the number of parameters or the amount of computation. Second , we propose atrous spatial pyramid pooling (ASPP) to robustly segment objects at multiple scales. ASPP probes an incoming convolutional feature layer with filters at multiple sampling rates and effective fields-of-views, thus capturing objects as well as image context at multiple scales. Third , we improve the localization of object boundaries by combining methods from DCNNs and probabilistic graphical models. The commonly deployed combination of max-pooling and downsampling in DCNNs achieves invariance but has a toll on localization accuracy. We overcome this by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF), which is shown both qualitatively and quantitatively to improve localization performance. Our proposed “DeepLab” system sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 79.7 percent mIOU in the test set, and advances the results on three other datasets: PASCAL-Context, PASCAL-Person-Part, and Cityscapes. All of our code is made publicly available online.

DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs

Image-to-image translation is a class of vision and graphics problems where the goal is to learn the mapping between an input image and an output image using a training set of aligned image pairs. However, for many tasks, paired training data will not be available. We present an approach for learning to translate an image from a source domain X to a target domain Y in the absence of paired examples. Our goal is to learn a mapping G : X → Y such that the distribution of images from G(X) is indistinguishable from the distribution Y using an adversarial loss. Because this mapping is highly under-constrained, we couple it with an inverse mapping F : Y → X and introduce a cycle consistency loss to push F(G(X)) ≈ X (and vice versa). Qualitative results are presented on several tasks where paired training data does not exist, including collection style transfer, object transfiguration, season transfer, photo enhancement, etc. Quantitative comparisons against several prior methods demonstrate the superiority of our approach.

Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

In this work we address the task of semantic image segmentation with Deep Learning and make three main contributions that are experimentally shown to have substantial practical merit. First, we highlight convolution with upsampled filters, or 'atrous convolution', as a powerful tool in dense prediction tasks. Atrous convolution allows us to explicitly control the resolution at which feature responses are computed within Deep Convolutional Neural Networks. It also allows us to effectively enlarge the field of view of filters to incorporate larger context without increasing the number of parameters or the amount of computation. Second, we propose atrous spatial pyramid pooling (ASPP) to robustly segment objects at multiple scales. ASPP probes an incoming convolutional feature layer with filters at multiple sampling rates and effective fields-of-views, thus capturing objects as well as image context at multiple scales. Third, we improve the localization of object boundaries by combining methods from DCNNs and probabilistic graphical models. The commonly deployed combination of max-pooling and downsampling in DCNNs achieves invariance but has a toll on localization accuracy. We overcome this by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF), which is shown both qualitatively and quantitatively to improve localization performance. Our proposed "DeepLab" system sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 79.7% mIOU in the test set, and advances the results on three other datasets: PASCAL-Context, PASCAL-Person-Part, and Cityscapes. All of our code is made publicly available online.

For a classification problem described by the joint density $P(\omega,x)$, models of $P(\omega\eq\omega'|x,x')$ (the ``Bayesian similarity measure'') have been shown to be an optimal similarity measure for nearest neighbor classification This paper analyzes demonstrates several additional properties of that conditional distribution The paper first shows that we can reconstruct, up to class labels, the class posterior distribution $P(\omega|x)$ given $P(\omega\eq\omega'|x,x')$, gives a procedure for recovering the class labels, and gives an asymptotically Bayes-optimal classification procedure It also shows, given such an optimal similarity measure, how to construct a classifier that outperforms the nearest neighbor classifier and achieves Bayes-optimal classification rates The paper then analyzes Bayesian similarity in a framework where a classifier faces a number of related classification tasks (multitask learning) and illustrates that reconstruction of the class posterior distribution is not possible in general Finally, the paper identifies a distinct class of classification problems using $P(\omega\eq\omega'|x,x')$ and shows that using $P(\omega\eq\omega'|x,x')$ to solve those problems is the Bayes optimal solution

On the Relationship between the Posterior and Optimal Similarity

Current symbol spotting and retrieval methods are not yet able to achieve the goal of both high accuracy and efficiency on large databases of line drawings. This paper presents an approach for focused symbol retrieval as step towards achieving such a goal by using concepts from image retrieval. During the off-line learning phase of the proposed approach, regions of interest are extracted from the drawings based on feature grouping. The regions are then described using an off-the-shelf descriptor. The similar descriptors are clustered, and finally a visual symbol vocabulary is learned by an SVM classifier. The vocabulary is constructed assuming no knowledge of the contents of the drawings. During on-line retrieval, the classifier recognizes the descriptors of query regions. A query can be a partial or a complete symbol, can contain contextual noise around a symbol or more than one symbol. Experimental results are presented for a database of architectural floor plans.

Towards Searchable Line Drawings, a Content-Based Symbol Retrieval Approach with Variable Query Complexity

L'invention concerne un systeme de recherches et de recommandations. Ce systeme utilise les preferences et les profils d'utilisateurs et de groupes individuels dans une communaute d'utilisateurs, ainsi que les informations derivees de signets de documents partages, pour augmenter les recherches d'Internet, reevaluer les resultats des recherches, et emettre des recommandations concernant ces documents sur la base d'interrogations sur des sujets specifiques. Le systeme de recherches et de recommandations fonctionne dans le contexte d'un gestionnaire de signets partages, qui memorise les signets des utilisateurs individuels (dont certains peuvent etre publies ou partages pour l'utilisation du groupe) sur une base de donnees de signets centralisee connectee a Internet. Le gestionnaire de signets partages est mis en oeuvre sous forme d'un programme reparti, dont certaines parties fonctionnent sur des terminaux d'utilisateurs et d'autres parties fonctionnent sur la base de donnees des signets centralises.

Systeme et procede de recherche et de recommandation de documents dans une collection a l'aide de signets partages

PROBLEM TO BE SOLVED: To provide a method for operating a white board printer capable of executing various mechanical works. SOLUTION: A forming sealing element or end effecter 130 for working on a substantially vertical surface such as a white board printer 105 or the like is held at or moved to its place by an effecter base 120. Since the base 120 is suspended by a left wire 114 and a right wire 112, the base 120 can be moved to any position between both the spools lower than a left spool 108 and a right spool 110. The method for moving the base 120 to the desired position by using the printer 105 is provided.

Method for operating on vertical surface

I describe an approach to similarity motivated by Bayesian methods. This yields a similarity function that is learnable using a standard Bayesian methods. The relationship of the approach to variable kernel and variable metric methods is discussed. The approach is related to variable kernel Experimental results on character recognition and 3D object recognition are presented.

Thomas M. Breuel

Papers

On the Relationship between the Posterior and Optimal Similarity

Towards Searchable Line Drawings, a Content-Based Symbol Retrieval Approach with Variable Query Complexity

Systeme et procede de recherche et de recommandation de documents dans une collection a l'aide de signets partages

Method for operating on vertical surface

Learning Similarity for Character Recognition and 3D Object Recognition