Semi-supervised learning, i.e. jointly learning from labeled and unlabeled samples, is an active research topic due to its key role on relaxing human supervision. In the context of image classification, recent advances to learn from unlabeled samples are mainly focused on consistency regularization methods that encourage invariant predictions for different perturbations of unlabeled samples. We, conversely, propose to learn from unlabeled data by generating soft pseudo-labels using the network predictions. We show that a naive pseudo-labeling overfits to incorrect pseudo-labels due to the so-called confirmation bias and demonstrate that mixup augmentation and setting a minimum number of labeled samples per mini-batch are effective regularization techniques for reducing it. The proposed approach achieves state-of-the-art results in CIFAR-10/100, SVHN, and Mini-ImageNet despite being much simpler than other methods. These results demonstrate that pseudo-labeling alone can outperform consistency regularization methods, while the opposite was supposed in previous work. Source code is available at https://git.io/fjQsC.

The prediction of salient areas in images has been traditionally addressed with hand-crafted features based on neuroscience principles. This paper, however, addresses the problem with a completely data-driven approach by training a convolutional neural network (convnet). The learning process is formulated as a minimization of a loss function that measures the Euclidean distance of the predicted saliency map with the provided ground truth. The recent publication of large datasets of saliency prediction has provided enough data to train end-to-end architectures that are both fast and accurate. Two designs are proposed: a shallow convnet trained from scratch, and a another deeper solution whose first three layers are adapted from another network trained for classification. To the authors' knowledge, these are the first end-to-end CNNs trained and tested for the purpose of saliency prediction.

This paper proposes a new approach to automatically quantify the severity of knee osteoarthritis (OA) from radiographs using deep convolutional neural networks (CNN). Clinically, knee OA severity is assessed using Kellgren & Lawrence (KL) grades, a five point scale. Previous work on automatically predicting KL grades from radiograph images were based on training shallow classifiers using a variety of hand engineered features. We demonstrate that classification accuracy can be significantly improved using deep convolutional neural network models pre-trained on ImageNet and fine-tuned on knee OA images. Furthermore, we argue that it is more appropriate to assess the accuracy of automatic knee OA severity predictions using a continuous distance-based evaluation metric like mean squared error than it is to use classification accuracy. This leads to the formulation of the prediction of KL grades as a regression problem and further improves accuracy. Results on a dataset of X-ray images and KL grades from the Osteoarthritis Initiative (OAI) show a sizable improvement over the current state-of-the-art.

In this paper, we discuss the criteria that should be satisfied by a descriptor for nonrigid shapes with a single closed contour. We then propose a shape representation method that fulfills these criteria. In the proposed approach, contour convexities and concavities at different scale levels are represented using a two-dimensional (2-D) matrix. The representation can be visualized as a 2-D surface, where "hills" and "valleys" represent contour convexities and concavities, respectively. The optimal matching of two shape representations is achieved using dynamic programming and a dissimilarity measure is defined based on this matching. The proposed algorithm is very efficient and invariant to several kinds of transformations including some articulations and modest occlusions. The retrieval performance of the approach is illustrated using the MPEG-7 shape database, which is one of the most complete shape databases currently available. Our experiments indicate that the proposed representation is well suited for object indexing and retrieval in large databases. Furthermore, the representation can be used as a starting point to obtain more compact descriptors.

In this paper, we propose a novel audio-visual feature-based framework for event detection in broadcast video of multiple different field sports. Features indicating significant events are selected and robust detectors built. These features are rooted in characteristics common to all genres of field sports. The evidence gathered by the feature detectors is combined by means of a support vector machine, which infers the occurrence of an event based on a model generated during a training phase. The system is tested generically across multiple genres of field sports including soccer, rugby, hockey, and Gaelic football and the results suggest that high event retrieval and content rejection statistics are achievable.