Abstract High dynamic range (HDR) imaging provides the capability of handling real world lighting as opposed to traditional low (LDR) which struggles accurately represent images with higher range. However, most content is still available only in LDR. This paper presents a method for generating HDR from LDR based on deep Convolutional Neural Networks (CNNs) termed ExpandNet. ExpandNet accepts input and generates an expanded end‐to‐end fashion. The model attempts reconstruct missing...

Dense captioning provides detailed captions of complex visual scenes. While a number successes have been achieved in recent years, there are still two broad limitations: 1) most existing methods adopt an encoder-decoder framework, where the contextual information is sequentially encoded using long short-term memory (LSTM). However, forget gate mechanism LSTM makes it vulnerable when dealing with sequence and 2) vast majority prior arts consider regions interests (RoIs) equally important,...

Dense captioning generates more detailed spoken descriptions for complex visual scenes. Despite several promising leads, existing methods still have two broad limitations: 1) The vast majority of prior arts only consider contextual clues during but ignore potentially important textual context; 2) current imbalanced learning mechanisms limit the diversity vocabulary learned from dictionary, thus giving rise to low language-learning efficiency. To alleviate these gaps, in this paper, we...

This is a repository copy of Virtual category learning: semi-supervised learning method for dense prediction with extremely limited labels.

Dense captioning creates diverse Region of Interests (RoIs) descriptions for complex visual scenes. While promising results have been obtained, several issues persist. In particular: 1) it is hard to find the optimal parameters artificially designed modules (e.g., non-maximum suppression (NMS)) causing redundancies and fewer interactions benefit two sub-tasks RoI detection captioning; 2) absence a multi-scale decoder in current methods hinders acquisition scale-invariant features, thus...

In recent years many Tone Mapping Operators (TMOs) have been presented in order to display High Dynamic Range Images (HDRI) on typical devices. TMOs compress the luminance range while trying maintain contrast. The dual of tone mapping, inverse expands a Low Image (LDRI) into HDRI. HDRIs contain broader physical values that can be perceived by human visual system. majority today's media is stored low dynamic range. Inverse (iTMOs) could thus potentially revive all this content for use high...

In recent years, there has been a notable surge in the adoption of weakly-supervised learning for medical image segmentation, utilizing scribble annotation as means to potentially reduce costs. However, inherent characteristics labeling, marked by incompleteness, subjectivity, and lack standardization, introduce inconsistencies into annotations. These become significant challenges network's process, ultimately affecting performance segmentation. To address this challenge, we propose creating...

Perceiving and understanding cyber-attacks can be a difficult task. This problem is widely recognized welldocumented, more effective techniques are needed to aid cyber-attack perception. Attack modeling (AMTs), such as attack graphs fault trees, useful visual aids that perception; however, there little empirical or comparative research which evaluates the effectiveness of these methods. paper reports results an evaluation between adapted graph method tree standard determine two methods in...

Deep learning-based semi-supervised learning (SSL) algorithms are promising in reducing the cost of manual annotation clinicians by using unlabelled data, when developing medical image segmentation tools. However, to date, most existing treat labelled images and separately ignore explicit connection between them; this disregards essential shared information thus hinders further performance improvements. To mine images, we introduce a class-specific representation extraction approach, which...

Increasing plastic recycling rates is key to addressing pollution. New technologies such as chemometric analysis of spectral data have shown great promises in improving the sorting efficiency boost rates. In this work, a novel deep learning architecture, PolymerSpectraDecisionNet (PSDN) was developed, consisting convolutional neural networks, residual networks and inception decision tree structure. To better represent conditions industry, models were built identify most widely recycled...

Abstract Sheet metal stamping is widely used for high-volume production. Despite the wide adoption, it can lead to defects in manufactured components, making their quality unacceptable. Because of variety that occur on final product, human inspectors are frequently employed detect them. However, they be unreliable and costly, particularly at speeds match rate. In this paper, we propose an automatic inspection framework process based computer vision deep learning techniques. The low cost,...

The computation of high-fidelity images in real-time remains one the key challenges for computer graphics. Recent work has shown that by understanding human visual system, selective rendering may be used to render only those parts which viewer is attending at high quality and rest scene a much lower quality. This can result significant reduction computational time, without being aware difference. Selective guided models typically form 2D saliency map, predict where user will looking any...

Abstract In the last few years, researchers in field of High Dynamic Range (HDR) Imaging have focused on providing tools for expanding Low (LDR) content generation HDR images due to growing popularity applications, such as photography and rendering via Image‐Based Lighting, imminent arrival displays consumer market. LDR expansion is required lack fast reliable level capture still videos. Furthermore, expansion, will allow re‐use legacy stills, videos applications created, over century more,...

Contemporary multi-modal trackers achieve strong performance by leveraging complex backbones and fusion strategies, but this comes at the cost of computational efficiency, limiting their deployment in resource-constrained settings. On other hand, compact are more efficient often suffer from reduced due to limited feature representation. To mitigate gap between trackers, we introduce a cross-modality distillation framework. This framework includes complementarity-aware mask autoencoder...