Automated pavement distresses detection using road images remains a challenging topic in the computer vision research community. Recent developments in deep learning has led to considerable research activity directed towards improving the efficacy of automated pavement distress identification and rating. Deep learning models require a large ground truth data set, which is often not readily available in the case of pavements. In this study, a labeled dataset approach is introduced as a first step towards a more robust, easy-to-deploy pavement condition assessment system. The technique is termed herein as the Pavement Image Dataset (PID) method. The dataset consists of images captured from two camera views of an identical pavement segment, i.e., a wide-view and a top-down view. The wide-view images were used to classify the distresses and to train the deep learning frameworks, while the top-down view images allowed calculation of distress density, which will be used in future studies aimed at automated pavement rating. For the wide view group dataset, 7,237 images were manually annotated and distresses classified into nine categories. Images were extracted using the Google Application Programming Interface (API), selecting street-view images using a python-based code developed for this project. The new dataset was evaluated using two mainstream deep learning frameworks: You Only Look Once (YOLO v2) and Faster Region Convolution Neural Network (Faster R-CNN). Accuracy scores using the F1 index were found to be 0.84 for YOLOv2 and 0.65 for the Faster R-CNN model runs; both quite acceptable considering the convenience of utilizing Google maps images.

For example, we might expect to encounter sensor malfunctions in a wireless sensor network at a rate proportional to the size of the network. Therefore, there is a growing need to develop machine learning techniques that enable satisfactory training and inference from incomplete data. Imputation, where missing data values are filled with suitable values inferred from observations, represents a promising technique for extending machine learning methods to handle missing data. Given their explicit representation of underlying data distributions, Restricted Boltzmann Machines (RBMs) are an appealing choice for imputing missing values. With a well trained RBM, the conditional probabilities of the missing values given the observed values remain accessible via either direct calculation (in a theoretical sense) or indirect Gibbs sampling. A variety of training and imputing procedures have been proposed to allow the application of RBMs to handle missing data, with various computational costs.

Compositional structures between parts and objects are inherent in natural scenes. Modeling such compositional hierarchies via unsupervised learning can bring various benefits such as interpretability and transferability, which are important in many downstream tasks. In this paper, we propose the first deep latent variable model, called RICH, for learning Representation of Interpretable Compositional Hierarchies. At the core of RICH is a latent scene graph representation that organizes the entities of a scene into a tree structure according to their compositional relationships. During inference, taking top-down approach, RICH is able to use higher-level representation to guide lower-level decomposition. This avoids the difficult problem of routing between parts and objects that is faced by bottom-up approaches. In experiments on images containing multiple objects with different part compositions, we demonstrate that RICH is able to learn the latent compositional hierarchy and generate imaginary scenes.

Neural networks (NNs) are able to perform tasks that rely on compositional structure even though they lack obvious mechanisms for representing this structure. To analyze the internal representations that enable such success, we propose ROLE, a technique that detects whether these representations implicitly encode symbolic structure. ROLE learns to approximate the representations of a target encoder E by learning a symbolic constituent structure and an embedding of that structure into E's representational vector space. The constituents of the approximating symbol structure are defined by structural positions --- roles --- that can be filled by symbols. We show that when E is constructed to explicitly embed a particular type of structure (string or tree), ROLE successfully extracts the ground-truth roles defining that structure. We then analyze a GRU seq2seq network trained to perform a more complex compositional task (SCAN), where there is no ground truth role scheme available. For this model, ROLE successfully discovers an interpretable symbolic structure that the model implicitly uses to perform the SCAN task, providing a comprehensive account of the representations that drive the behavior of a frequently-used but hard-to-interpret type of model. We verify the causal importance of the discovered symbolic structure by showing that, when we systematically manipulate hidden embeddings based on this symbolic structure, the model's resulting output is changed in the way predicted by our analysis. Finally, we use ROLE to explore whether popular sentence embedding models are capturing compositional structure and find evidence that they are not; we conclude by discussing how insights from ROLE can be used to impart new inductive biases to improve the compositional abilities of such models.

This paper addresses a boosting method for mapping functionality of neural networks in visual recognition such as image classification and face recognition. We present reversible learning for generating and learning latent features using the network itself. By generating latent features corresponding to hard samples and applying the generated features in a training stage, reversible learning can improve a mapping functionality without additional data augmentation or handling the bias of dataset. We demonstrate an efficiency of the proposed method on the MNIST,Cifar-10/100, and Extremely Biased and poorly categorized dataset (EBPC dataset). The experimental results show that the proposed method can outperform existing state-of-the-art methods in visual recognition. Extensive analysis shows that our method can efficiently improve the mapping capability of a network.

Real artificial intelligence always has been focused on by many machine learning researchers, especially in the area of deep learning. However deep neural network is hard to be understood and explained, and sometimes, even metaphysics. The reason is, we believe that: the network is essentially a perceptual model. Therefore, we believe that in order to complete complex intelligent activities from simple perception, it is necessary to con-struct another interpretable logical network to form accurate and reasonable responses and explanations to external things. Researchers like Bolei Zhou and Quanshi Zhang have found many explanatory rules for deep feature extraction aimed at the feature extraction stage of convolution neural network. However, although researchers like Marco Gori have also made great efforts to improve the interpretability of the fully connected layers of the network, the problem is also very difficult. This paper firstly analyzes its reason. Then a method of constructing logical network based on the fully connected layers and extracting logical relation between input and output of the layers is proposed. The game process between perceptual learning and logical abstract cognitive learning is implemented to improve the interpretable performance of deep learning process and deep learning model. The benefits of our approach are illustrated on benchmark data sets and in real-world experiments.

The interpretation of black-box models has been investigated in recent years. A number of model-aware saliency methods were proposed to explain individual classification decisions by creating saliency maps. However, they are not applicable when the parameters and the gradients of the underlying models are unavailable. Recently, model-agnostic methods have received increased attention. As one of them, Prediction Difference Analysis (PDA), a probabilistic sound methodology, was proposed. In this work, we first show that PDA can suffer from saturated classifiers. The saturation phenomenon of classifiers exists widely in current neural network-based classifiers. To understand the decisions of saturated classifiers better, we further propose Contextual PDA, which runs hundreds of times faster than PDA. The experiments show the superiority of our method by explaining image classifications of the state-of-the-art deep convolutional neural networks. We also apply our method to commercial general vision recognition systems.

Classification problems today are typically solved by first collecting examples along with candidate labels, second obtaining clean labels from workers, and third training a large, overparameterized deep neural network on the clean examples. The second, labeling step is often the most expensive one as it requires manually going through all examples. In this paper we skip the labeling step entirely and propose to directly train the deep neural network on the noisy candidate labels and early stop the training to avoid overfitting. With this procedure we exploit an intriguing property of large overparameterized neural networks: While they are capable of perfectly fitting the noisy data, gradient descent fits clean labels much faster than the noisy ones, thus early stopping resembles training on the clean labels. Our results show that early stopping the training of standard deep networks such as ResNet-18 on part of the Tiny Images dataset, which does not involve any human labeled data, and of which only about half of the labels are correct, gives a significantly higher test performance than when trained on the clean CIFAR-10 training dataset, which is a labeled version of the Tiny Images dataset, for the same classification problem. In addition, our results show that the noise generated through the label collection process is not nearly as adversarial for learning as the noise generated by randomly flipping labels, which is the noise most prevalent in works demonstrating noise robustness of neural networks.

This paper proposes a new evaluation metric and boosting method for weight separability in neural network design. In contrast to general visual recognition methods designed to encourage both intra-class compactness and inter-class separability of latent features, we focus on estimating linear independence of column vectors in weight matrix and improving the separability of weight vectors. To this end, we propose an evaluation metric for weight separability based on semi-orthogonality of a matrix and Frobenius distance, and the feed-backward reconstruction loss which explicitly encourages weight separability between the column vectors in the weight matrix. The experimental results on image classification and face recognition demonstrate that the weight separability boosting via minimization of feed-backward reconstruction loss can improve the visual recognition performance, hence universally boosting the performance on various visual recognition tasks.

Incorporating human expertise and domain knowledge is particularly important for medical image processing applications, marked with small datasets, and objects of interests in the form of organs or lesions not typically seen in traditional datasets. However, the incorporation of prior knowledge for breast tumor detection is challenging, since shape, boundary, curvature, intensity, or other common medical priors vary significantly across patients and cannot be employed. This work proposes an approach for integrating visual saliency into a deep learning model for breast tumor segmentation in ultrasound images. Visual saliency emphasizes regions that are more likely to attract radiologists' visual attention and stand out from its surrounding. Our approach is based on a U-Net model and employs attention blocks to introduce visual saliency. Such model forces learning feature representations that prioritize spatial regions with high levels of saliency. The approach is validated using a dataset of 510 breast ultrasound images.