In this work, we pose the question whether it is possible to design and train an autoencoder model in an end-to-end fashion to learn representations in the multivariate Bernoulli latent space, and achieve performance comparable with the state-of-the-art variational methods. Moreover, we investigate how to generate novel samples and perform smooth interpolation and attributes modification in the binary latent space. To meet our objective, we propose a simplified, deterministic model with a straight-through gradient estimator to learn the binary latents and show its competitiveness with the latest VAE methods. Furthermore, we propose a novel method based on a random hyperplane rounding for sampling and smooth interpolation in the latent space. Our method performs on a par or better than the current state-of-the-art methods on common CelebA, CIFAR-10 and MNIST datasets.

Within this work a novel semi-supervised learning technique is introduced based on a simple iterative learning cycle together with learned thresholding techniques and an ensemble decision support system. State-of-the-art model performance and increased training data volume are demonstrated, through the use of unlabelled data when training deeply learned classification models. The methods presented work independently from the model architectures or loss functions, making this approach applicable to a wide range of machine learning and classification tasks. Evaluation of the proposed approach is performed on commonly used datasets when evaluating semi-supervised learning techniques as well as a number of more challenging image classification datasets (CIFAR-100 and a 200 class subset of ImageNet).

In this paper we present the design and evaluation of an end-to-end trainable, deep neural network with a visual attention mechanism for memorability estimation in still images. We analyze the suitability of transfer learning of deep models from image classification to the memorability task. Further on we study the impact of the attention mechanism on the memorability estimation and evaluate our network on the SUN Memorability and the LaMem datasets. Our network outperforms the existing state of the art models on both datasets in terms of the Spearman's rank correlation as well as the mean squared error, closely matching human consistency.

In this work, we extract rich representations of crowd behavior from video using a fine-tuned deep convolutional neural residual network. Using spatial partitioning trees we create subclasses within the feature maps from each of the crowd behavior attributes (classes). Features from these subclasses are then regularized using an eigen modeling scheme. This enables to model the variance appearing from the intra-subclass information. Low dimensional discriminative features are extracted after using the total subclass scatter information. Dynamic time warping is used on the cosine distance measure to find the similarity measure between videos. A 1-nearest neighbor (NN) classifier is used to find the respective crowd behavior attribute classes from the normal videos. Experimental results on large crowd behavior video database show the superior performance of our proposed framework as compared to the baseline and current state-of-the-art methodologies for the crowd behavior recognition task.