Showing papers in "Computer Vision and Image Understanding in 2019"

TL;DR: More recently, this article reviewed and critically discussed more than 24 quantitative and 5 qualitative measures for evaluating generative models with a particular emphasis on GAN-derived models and also provided a set of 7 desiderata followed by an evaluation of whether a given measure or a family of measures is compatible with them.

TL;DR: Major deep learning concepts pertinent to face image analysis and face recognition are reviewed, and a concise overview of studies on specific face recognition problems is provided, such as handling variations in pose, age, illumination, expression, and heterogeneous face matching.

TL;DR: This paper proposes a general end-to-end network, called the Anabranch Network, that leverages both classification and segmentation tasks and possesses the second branch for classification to predict the probability of containing camouflaged object(s) in an image.

TL;DR: The Exclusively Dark dataset as discussed by the authors consists of low-light images captured in visible light only, with image and object level annotations, and the effects of lowlight reach far deeper into the features than can be solved by simple illumination invariance.

TL;DR: This paper presents the first large scale very high resolution semantic change detection dataset, which enables the usage of deep supervised learning methods for semantic changes detection with very highresolution images, and presents a network architecture that performs change detection and land cover mapping simultaneously.

TL;DR: The detail analysis of different V-reID methods in terms of mean average precision (mAP) and cumulative matching curve (CMC) provide objective insight into the strengths and weaknesses of these methods.

TL;DR: This paper proposes the SGCN architecture for assessing the similarity between a pair of graphs which can be trained with the contrastive loss function and implements the proposed embeddings for the task of CBIR for RS data on the popular UC-Merced dataset and the PatternNet dataset where improved performance can be observed.

TL;DR: A proposed fully-unsupervised machine learning algorithm converts the radar sensor data to artificial, camera-like, environmental images that are more consistent, accurate, and useful information than that provided solely by the radar or the camera.

TL;DR: In this paper, a survey provides a summary of these challenges and datasets to address them, followed by an in-depth discussion of relevant vision-based recognition and detection methods, focusing on recent, promising work based on deep learning and convolutional neural networks (CNNs).

TL;DR: Zhang et al. as discussed by the authors proposed a new deep neural network for object detection, which builds feature relations in the spatial space of the feature map, and learns to highlight useful regions on the feature maps while suppressing the irrelevant information.

TL;DR: This work uses the natural structure of the hand topology – called later hand skeletal data – to extract effective hand kinematic descriptors from the gesture sequence and introduces a prior gesture detection phase achieved using a binary classifier before the final gesture recognition.

TL;DR: Guo et al. as discussed by the authors designed a multi-task leaning architecture in an end-to-end manner to reduce the mapping range from input to output and boost the performance, where a decomposition net is built to split rain images into clean background and rain layers.

TL;DR: Face-SSD is the first network to perform face analysis without relying on pre-processing such as face detection and registration in advance and achieves real-time performance even when detecting multiple faces and recognising multiple classes in a given image.

TL;DR: An auxiliary task to Mask R-CNN, an instance segmentation network, is presented, which leads to faster training of the mask head, and a new prediction head is added, the Edge Agreement Head, which is inspired by the way human annotators perform instance segmentations.

TL;DR: 3DDE as discussed by the authors is a robust and efficient face alignment algorithm based on a coarse-to-fine cascade of ensembles of regression trees, which is initialized by robustly fitting a 3D face model to the probability maps produced by a CNN.

TL;DR: In this article, a semi-supervised regression GAN is proposed for regression problems and applied to age estimation, driving steering angle prediction, and crowd counting from a single image.

TL;DR: The proposed alternative framework where the cost function used for inferring a parametric transfer function is defined as the robust L 2 divergence between two probability density functions outperforms many recent algorithms as measured quantitatively with standard quality metrics, and qualitatively using perceptual studies.

TL;DR: In this paper, the authors address the problem of generating images across two drastically different views, namely ground (street) and aerial (overhead) views, and resort to homography as a guide to map the images between the views based on the common field of view to preserve the details in the input image.

TL;DR: Li et al. as mentioned in this paper proposed a unified framework which jointly solves both person re-identification and camera network topology inference problems with minimal prior knowledge about the environments, which can be applied to online person Re-ID in large-scale multi-camera networks.

TL;DR: This work introduces a new semantic segmentation regularization based on the regression of a distance transform, which requires almost no modification of the network structure and adds a very low overhead to the training process.

TL;DR: This paper suggests an adaptation of the random forest classifier by integrating a model for label noise based on the idea that a training sample should not be assigned to one class only, but to all classes, each with a certain probability.

TL;DR: This paper utilizes a variant of Self-organizing Map to cluster images in two different scenarios: disjoint and non-disjoint sets, and compares the state-of-the-art image clustering algorithms with a SOM-based subspace clustering method that identifies automatically the relevant features in the high-dimensional image representations.

TL;DR: This study is the first review of published video synopsis approaches and provides a comprehensive analysis of state-of-the-art approaches to achieve efficient video browsing and retrieval for surveillance cameras.

TL;DR: This approach combines geometric smoothness priors in the image space with more traditional uncertainty measures to estimate which pixels or voxels are the most informative, and thus should to be annotated next, for multi-class settings and introduces two novel criteria for uncertainty.

TL;DR: This paper proposes a recurrent attention mechanism for VQA which utilizes dual (textual and visual) Recurrent Attention Units (RAUs) and shows the effect of all possible combinations of recurrent and convolutional dual attention.

TL;DR: In this article, an active learning framework is proposed to select most informative samples for training a robust deep learning system using a Bayesian neural network, which is then used within a novel class aware generative adversarial network (CAGAN) to generate realistic chest xray images for data augmentation by transferring characteristics from one class label to another.

TL;DR: In this article, an end-to-end architecture comprising a GAN to address the domain shift at training time and a deep CNN trained on the samples generated by the GAN was proposed to learn an embedding of product images.

TL;DR: Compared to existing models, synthetic face images generated by the proposed attribute guided face image generation method present a good photorealistic quality on several face datasets and can be used for synthetic data augmentation, and improve the performance of the classifier used for facial expression recognition.

TL;DR: This work proposes an approach based on a Conditional Generative Adversarial Network (CGAN) for refining the coarse reconstruction provided by a 3DMM, represented as a three channels image, where the pixel intensities represent the depth, curvature and elevation values of the 3D vertices.

TL;DR: This paper presents an effective framework named Attentive Matching Network (AMN) to address few-shot learning problem, and proposes a feature-level attention mechanism to help similarity function pay more emphasis on the features that better reflect the inter-class differences.