Joint Learning of Cross-Modal Attribute and Semantic Representation for Action Recognition – The proposed algorithm is a novel deep neural network architecture for collaborative object detection in natural images. A key task of this framework is to find an object belonging to the object category in a given image, and the classification of the object can be performed on a class label for each image, which is then used to detect the object type. Despite its simplicity, a deep learning based approach is essential for an effective and effective method for this purpose. We present the first deep learning based approach for collaborative object detection in an unsupervised manner which can be used in a variety of applications from image search to image understanding. Extensive evaluations on various benchmark datasets, including Flickr30K in both computer vision and image processing, show that the proposed deep learning framework achieves comparable or superior performance with respect to state-of-the-art object detection methods in terms of both accuracy and recall.

The recent emergence of deep learning has brought many new challenges to human-computer interaction including the problem of designing and learning artificial agents. In fact, the task of real-time interaction with agents is important for a wide range of applications including interactive and game-like games. We tackle the problem in two steps. First, we model the world as a set of interconnected objects (i.e., agents) that interact with each other as a natural learning mechanism. Second, we learn agents that exploit the system resources. As with any learning system, a new agent could or could not be learned yet. To explore the potential for agents we apply deep learning techniques to the world of action and action recognition. We show that deep learning can help agents to better understand, perform and explore the system resources efficiently.

Matching-Gap Boosting for Sparse Signal Recovery with Nonparametric Noise Distributions

Tensor-based transfer learning for image recognition

Joint Learning of Cross-Modal Attribute and Semantic Representation for Action Recognition

Predicting Daily Activity with a Deep Neural Network

Fractal-based Deep Convolutional Representations: Algorithms and ComparisonsThe recent emergence of deep learning has brought many new challenges to human-computer interaction including the problem of designing and learning artificial agents. In fact, the task of real-time interaction with agents is important for a wide range of applications including interactive and game-like games. We tackle the problem in two steps. First, we model the world as a set of interconnected objects (i.e., agents) that interact with each other as a natural learning mechanism. Second, we learn agents that exploit the system resources. As with any learning system, a new agent could or could not be learned yet. To explore the potential for agents we apply deep learning techniques to the world of action and action recognition. We show that deep learning can help agents to better understand, perform and explore the system resources efficiently.