Weakly supervised road network extraction for remote sensing image based scribble annotation and adversarial learning

Yuan, Genji; Li, Jianbo; Li, Xue; Zheng, Yufeng

doi:10.1016/j.jksuci.2022.05.020

Cited by 13 publications

(4 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With the rapid development of deep learning in various fields [4,[33][34][35], a large number of deep learning-based image processing algorithms [36,37] have been proposed, and substantial progress has been made.…”

Section: Data-driven Dehazing Methodsmentioning

confidence: 99%

Multi-level perception fusion dehazing network

Wu,

Li,

Guo

et al. 2023

PLoS ONE

View full text Add to dashboard Cite

Image dehazing models are critical in improving the recognition and classification capabilities of image-related artificial intelligence systems. However, existing methods often ignore the limitations of receptive field size during feature extraction and the loss of important information during network sampling, resulting in incomplete or structurally flawed dehazing outcomes. To address these challenges, we propose a multi-level perception fusion dehazing network (MPFDN) that effectively integrates feature information across different scales, expands the perceptual field of the network, and fully extracts the spatial background information of the image. Moreover, we employ an error feedback mechanism and a feature compensator to address the loss of features during the image dehazing process. Finally, we subtract the original hazy image from the generated residual image to obtain a high-quality dehazed image. Based on extensive experimentation, our proposed method has demonstrated outstanding performance not only on synthesizing dehazing datasets, but also on non-homogeneous haze datasets.

show abstract

Section: Data-driven Dehazing Methodsmentioning

confidence: 99%

Multi-level perception fusion dehazing network

Wu,

Li,

Guo

et al. 2023

PLoS ONE

View full text Add to dashboard Cite

show abstract

“…The regression process can estimate values below the minimum threshold (zero) and above the maximum threshold (five) due to the absence of constraints. Wang et al [55] introduce the semantic knowledge mapping network (S-KMN) to improve quiz question annotation. It combines semantic feature learning and knowledge mapping, addressing limitations in existing studies.…”

Section: ) Techniques For Text Representationmentioning

confidence: 99%

A Review on NLP Techniques and Associated Challenges in Extracting Features from Education Data

Ahidi Elisante Lukwaro,

Kalegele,

G. Nyambo

2024

IJCDS

View full text Add to dashboard Cite

There has been a significant increase in academic processes to ensure the quality of educational resources such as curricula, examinations, and educational content. This has drawn attention to studies exploring the use of text mining, learning machines, and auto-analytic tools like natural language processing (NLP) to interpret and evaluate the quality of these educational resources. Auto-analytical techniques are required to evaluate the quality of educational content; otherwise, manual evaluation can be burdensome and improperly influenced by human instincts. This study employs a methodical approach to comprehensively survey NLP techniques for extracting syntactic and semantic features to analyze and comprehend educational content. NLP, in combination with machine learning, is an ideal tool for automatically evaluating the aspects of higher education quality. This is because they include features that aid in textual content comprehension as well as implementing natural language techniques that provide an interpretive interface between humans and machines. The review highlights the limitations of NLP in evaluating educational data, including the need for sentence-level understanding and the need for research to address challenges like noise in text data, domain-specific language variations, and improving model robustness for effective feature extraction in educational contexts. The findings of this review hold substantial benefits for various stakeholders, including education regulatory bodies, researchers, higher education institutions, and NLP researchers. Notably, the study equips NLP researchers with valuable insights into document analysis's current strengths and weaknesses. The accumulated evidence can provide the skills to develop NLP-based applications for evaluating the relevant and quality aspects of education in higher educational settings. Furthermore, NLP researchers can be updated on the strengths and limitations of document analysis, allowing them to apply effective text representation approaches and implement the appropriate algorithm and techniques for NLP tasks, particularly in educational data.

show abstract

“…This study uses both Kappa consistency test and F1 Score [38] to verify the accuracy of land use forecast data in 2025. 5).…”

Section: Model Accuracy Verificationmentioning

confidence: 99%

Spatiotemporal Evolution of Ecosystem Services in the Wanhe Watershed Based on Cellular Automata (CA)-Markov and InVEST Models

2022

View full text Add to dashboard Cite

The evaluation of habitat quality and its genesis is of great significance to ecological protection of the watershed. Based on land use data, Digital Elevation Model (DEM), and road network data and population data, the Cellular Automata (CA)-Markov model and InVEST model were used to analyzed the land use change in the Wanhe Watershed, predicting the land use in 2025. Based on this, the degree of the habitat degradation and habitat quality in 2000, 2005, 2010, 2015, 2020, and 2025 were predicted and analyzed, and combined with the particularity of the terrain in the study area, the topography was introduced. Landform relief gradient was used to discuss the relationship between habitat quality and topographic factors in the Wanhe Watershed, and to reveal the distribution law. The result shows that from 2000 to 2025, farmland and forestland are the main land use types in the study area, and the main change is due to the expansion of the construction land, whereby the area increased by 62.86 km2, with an increase of 34.41%, mainly from farmland and forestland. From 2000 to 2020, mainly due to the expansion in urban land and the reduction in forestland, the habitat degradation tends to be serious, and the habitat quality generally shows a downward trend, with areas with low habitat quality having had the largest increase from 452.67 km² in 2000 to 526.15 km² in 2025. The topography of the study area affects the distribution of natural landscapes and the intensity of human activities, resulting in significant differences in the landscape pattern of habitat degradation and habitat quality. The western mountains are relatively better. After 2020, due to the implementation of environmental protection policies, the habitat quality has tended to improve. This study can promote the adjustment of land use planning policies in the study area, maintain the biodiversity in the watershed, and realize the coordinated development of environmental benefits and social and economic development. The research results have theoretical significance and practical value for ecological environmental protection and land use layout in Wanhe Watershed.

show abstract

Weakly supervised road network extraction for remote sensing image based scribble annotation and adversarial learning

Cited by 13 publications

References 28 publications

Multi-level perception fusion dehazing network

Multi-level perception fusion dehazing network

A Review on NLP Techniques and Associated Challenges in Extracting Features from Education Data

Spatiotemporal Evolution of Ecosystem Services in the Wanhe Watershed Based on Cellular Automata (CA)-Markov and InVEST Models

Contact Info

Product

Resources

About