Single Image Dehazing Using Sparse Contextual Representation

In this paper, we propose a novel method to remove haze from a single hazy input image based on the sparse representation. In our method, the sparse representation is proposed to be used as a contextual regularization tool, which can reduce the block artifacts and halos produced by only using dark channel prior without soft matting as the transmission is not always constant in a local patch. A novel way to use dictionary is proposed to smooth an image and generate the sharp dehazed result. Experimental results demonstrate that our proposed method performs favorably against the state-of-the-art dehazing methods and produces high-quality dehazed and vivid color results.

Are Landscape Restoration Interventions Sustainable? A Case for the Upper Tana Basin of Kenya

Landscape restoration initiatives often have the potential to result in environmental gains, but the question of whether these gains are sustainable and how they are linked to other community needs (social, productivity and economic gains) remains unclear. We use the Sustainable Intensification Assessment Framework (SIAF) to demonstrate how environmental benefits are linked to productivity, environment, social, human, and economic components. Using the SIAF, the standardization of relevant indicators across multiple objectives provided a contextual representation of sustainability. The study assessed the overall gains resulting from the measured indicators of sustainable land management (SLM) practices and their relationship to the multiple domains of the SIAF. We present a unique case for SLM options using a combined-methods approach where biophysical, socio-economic, and citizen science help assess the sustainability of the interventions. Using a participatory approach with farmers, land restoration options were conducted in four target micro-watersheds for 3 years (2015–2017). Co-developed restoration measures at the landscape level within the four micro-watersheds (MW1-MW4) resulted in a substantial increment (50%) for all treatments (grass strips, terraces, and a combination of grass strips and terraces) in soil moisture storage and increased maize and forage production. We demonstrate that SLM practices, when used in combination, greatly reduce soil erosion and are profitable and sustainable while conferring livelihood benefits to smallholder farmers.

Hybridizing Cross-Level Contextual and Attentive Representations for Remote Sensing Imagery Semantic Segmentation

Semantic segmentation of remote sensing imagery is a fundamental task in intelligent interpretation. Since deep convolutional neural networks (DCNNs) performed considerable insight in learning implicit representations from data, numerous works in recent years have transferred the DCNN-based model to remote sensing data analysis. However, the wide-range observation areas, complex and diverse objects and illumination and imaging angle influence the pixels easily confused, leading to undesirable results. Therefore, a remote sensing imagery semantic segmentation neural network, named HCANet, is proposed to generate representative and discriminative representations for dense predictions. HCANet hybridizes cross-level contextual and attentive representations to emphasize the distinguishability of learned features. First of all, a cross-level contextual representation module (CCRM) is devised to exploit and harness the superpixel contextual information. Moreover, a hybrid representation enhancement module (HREM) is designed to fuse cross-level contextual and self-attentive representations flexibly. Furthermore, the decoder incorporates DUpsampling operation to boost the efficiency losslessly. The extensive experiments are implemented on the Vaihingen and Potsdam benchmarks. In addition, the results indicate that HCANet achieves excellent performance on overall accuracy and mean intersection over union. In addition, the ablation study further verifies the superiority of CCRM.

Pedagogically Sound Learning Objects: Towards a Useful Classification

In spite of the numerous discussions in literature, the learning object remains an illdefined concept. In this paper, rather than attempting to clearly define what a learning object is, I discuss kinds of computer-based creations that might be recognized as a learning object by the community involved in design and use of technology-based educational resources. This discussion is supported by a small-scale inquiry into kinds of learning objects identified from a collection of resources developed by some teachers and instructional designers in Singapore. Six unique categories of potential learning objects were noted and defined through the inquiry: presentation object, practice object, information object, simulation object, conceptual model and contextual representation. These kinds of learning objects are discussed in this paper. The paper opens a possibility for the proposed categories to be challenged or for more categories of learning objects to emerge in further inquiries involving examination of larger repositories of learning objects.

Pipeline corrosion defect parameterisation with magnetic flux leakage inspection: a contextual representation approach

Corrosion is one of the significant reasons for oil and gas pipeline failures. In pipeline integrity management programmes, the magnetic flux leakage (MFL) technique is widely used to detect and quantify corrosion defects. The inspection results of MFL list the profiles of individual corrosion defects; however, the structural safety of a pipeline not only depends on the size of individual corrosion defects but also the pattern of closely spaced defects. To achieve a contextual defect representation, the concept of parameterisation, which considers the adjacent defects as additional information of the central defect, is proposed in this study. The process through which to realise this contextual representation is described in this paper. Three parameterisation models are proposed and a two-dimensional Gaussian function is employed to model the interaction strength between adjacent defects. The experimental results demonstrate that the shape context (SC) model associated with the interaction strength function (ISF) shares the highest similarity with human inspectors in comparison with the other two models. Thus, the proposed parameterisation approach can be used to retrieve similar corrosion defects and analyse defect population distribution along a pipeline.

BERTMeSH: deep contextual representation learning for large-scale high-performance MeSH indexing with full text

Motivation With the rapid increase of biomedical articles, large-scale automatic Medical Subject Headings (MeSH) indexing has become increasingly important. FullMeSH, the only method for large-scale MeSH indexing with full text, suffers from three major drawbacks: FullMeSH (i) uses Learning To Rank, which is time-consuming, (ii) can capture some pre-defined sections only in full text and (iii) ignores the whole MEDLINE database. Results We propose a computationally lighter, full text and deep-learning-based MeSH indexing method, BERTMeSH, which is flexible for section organization in full text. BERTMeSH has two technologies: (i) the state-of-the-art pre-trained deep contextual representation, Bidirectional Encoder Representations from Transformers (BERT), which makes BERTMeSH capture deep semantics of full text. (ii) A transfer learning strategy for using both full text in PubMed Central (PMC) and title and abstract (only and no full text) in MEDLINE, to take advantages of both. In our experiments, BERTMeSH was pre-trained with 3 million MEDLINE citations and trained on ∼1.5 million full texts in PMC. BERTMeSH outperformed various cutting-edge baselines. For example, for 20 K test articles of PMC, BERTMeSH achieved a Micro F-measure of 69.2%, which was 6.3% higher than FullMeSH with the difference being statistically significant. Also prediction of 20 K test articles needed 5 min by BERTMeSH, while it took more than 10 h by FullMeSH, proving the computational efficiency of BERTMeSH. Supplementary information Supplementary data are available at Bioinformatics online