scholarly journals SEMI-SUPERVISED SEGMENTATION OF CONCRETE AGGREGATE USING CONSENSUS REGULARISATION AND PRIOR GUIDANCE

2021 ◽  
Vol V-2-2021 ◽  
pp. 83-91
Author(s):  
M. Coenen ◽  
T. Schack ◽  
D. Beyer ◽  
C. Heipke ◽  
M. Haist
Keyword(s):  
Prior Knowledge ◽  
Prior Information ◽  
Semantic Segmentation ◽  
Concrete Aggregate ◽  
Class Distribution ◽  
Supervised Segmentation ◽  
Additional Branch ◽  
Unlabelled Data ◽  
Additional Losses ◽  
Additional Constraints

Abstract. In order to leverage and profit from unlabelled data, semi-supervised frameworks for semantic segmentation based on consistency training have been proven to be powerful tools to significantly improve the performance of purely supervised segmentation learning. However, the consensus principle behind consistency training has at least one drawback, which we identify in this paper: imbalanced label distributions within the data. To overcome the limitations of standard consistency training, we propose a novel semi-supervised framework for semantic segmentation, introducing additional losses based on prior knowledge. Specifically, we propose a lightweight architecture consisting of a shared encoder and a main decoder, which is trained in a supervised manner. An auxiliary decoder is added as additional branch in order to make use of unlabelled data based on consensus training, and we add additional constraints derived from prior information on the class distribution and on auto-encoder regularisation. Experiments performed on our concrete aggregate dataset presented in this paper demonstrate the effectiveness of the proposed approach, outperforming the segmentation results achieved by purely supervised segmentation and standard consistency training.

scholarly journals Knowledge and Geo-Object Based Graph Convolutional Network for Remote Sensing Semantic Segmentation

Sensors ◽  
2021 ◽  
Vol 21 (11) ◽  
pp. 3848
Author(s):  
Wei Cui ◽  
Meng Yao ◽  
Yuanjie Hao ◽  
Ziwei Wang ◽  
Xin He ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Prior Knowledge ◽  
Contextual Information ◽  
Information Aggregation ◽  
Semantic Segmentation ◽  
Spatial Correlations ◽  
Convolutional Network ◽  
Object Based ◽  
Graph Neural Networks ◽  
Salt And Pepper

Pixel-based semantic segmentation models fail to effectively express geographic objects and their topological relationships. Therefore, in semantic segmentation of remote sensing images, these models fail to avoid salt-and-pepper effects and cannot achieve high accuracy either. To solve these problems, object-based models such as graph neural networks (GNNs) are considered. However, traditional GNNs directly use similarity or spatial correlations between nodes to aggregate nodes’ information, which rely too much on the contextual information of the sample. The contextual information of the sample is often distorted, which results in a reduction in the node classification accuracy. To solve this problem, a knowledge and geo-object-based graph convolutional network (KGGCN) is proposed. The KGGCN uses superpixel blocks as nodes of the graph network and combines prior knowledge with spatial correlations during information aggregation. By incorporating the prior knowledge obtained from all samples of the study area, the receptive field of the node is extended from its sample context to the study area. Thus, the distortion of the sample context is overcome effectively. Experiments demonstrate that our model is improved by 3.7% compared with the baseline model named Cluster GCN and 4.1% compared with U-Net.

Optimal Scale of Hierarchical Image Segmentation with Scribbles Guidance for Weakly Supervised Semantic Segmentation

2021 ◽  
pp. 2154026
Author(s):  
Zaid Al-Huda ◽  
Donghai Zhai ◽  
Yan Yang ◽  
Riyadh Nazar Ali Algburi
Keyword(s):  
Image Segmentation ◽  
Graphical Model ◽  
Semantic Segmentation ◽  
Saliency Map ◽  
Training Data ◽  
Deep Convolutional Neural Networks ◽  
High Quality ◽  
Optimal Scale ◽  
Supervised Segmentation ◽  
Weakly Supervised

Deep convolutional neural networks (DCNNs) trained on the pixel-level annotated images have achieved improvements in semantic segmentation. Due to the high cost of labeling training data, their applications may have great limitation. However, weakly supervised segmentation approaches can significantly reduce human labeling efforts. In this paper, we introduce a new framework to generate high-quality initial pixel-level annotations. By using a hierarchical image segmentation algorithm to predict the boundary map, we select the optimal scale of high-quality hierarchies. In the initialization step, scribble annotations and the saliency map are combined to construct a graphic model over the optimal scale segmentation. By solving the minimal cut problem, it can spread information from scribbles to unmarked regions. In the training process, the segmentation network is trained by using the initial pixel-level annotations. To iteratively optimize the segmentation, we use a graphical model to refine segmentation masks and retrain the segmentation network to get more precise pixel-level annotations. The experimental results on Pascal VOC 2012 dataset demonstrate that the proposed framework outperforms most of weakly supervised semantic segmentation methods and achieves the state-of-the-art performance, which is [Formula: see text] mIoU.

scholarly journals Edge Prior Multilayer Segmentation Network Based on Bayesian Framework

2020 ◽  
Vol 2020 ◽  
pp. 1-11 ◽  
Author(s):  
Chu He ◽  
Zishan Shi ◽  
Peizhang Fang ◽  
Dehui Xiong ◽  
Bokun He ◽  
...  
Keyword(s):  
Prior Information ◽  
Semantic Segmentation ◽  
Bayesian Framework ◽  
Excellent Performance ◽  
Transform Method ◽  
Data Set ◽  
Distance Map ◽  
Constraint Network ◽  
Average Accuracy ◽  
Complex Boundaries

In recent years, methods based on neural network have achieved excellent performance for image segmentation. However, segmentation around the edge area is still unsatisfactory when dealing with complex boundaries. This paper proposes an edge prior semantic segmentation architecture based on Bayesian framework. The entire framework is composed of three network structures, a likelihood network and an edge prior network at the front, followed by a constraint network. The likelihood network produces a rough segmentation result, which is later optimized by edge prior information, including the edge map and the edge distance. For the constraint network, the modified domain transform method is proposed, in which the diffusion direction is revised through the newly defined distance map and some added constraint conditions. Experiments about the proposed approach and several contrastive methods show that our proposed method had good performance and outperformed FCN in terms of average accuracy for 0.0209 on ESAR data set.

Reducing sample size through the use of a composite estimator: an application to timber volume estimation

1986 ◽  
Vol 16 (5) ◽  
pp. 1116-1118 ◽  
Author(s):  
Edwin J. Green ◽  
William E. Strawderman
Keyword(s):  
Sample Size ◽  
Prior Knowledge ◽  
Prior Information ◽  
Volume Estimation ◽  
Timber Volume ◽  
Percent Error ◽  
Sample Information ◽  
Sample Data ◽  
Composite Estimator

A method for determining the appropriate sample size to produce an estimate with a stated allowable percent error when the sample data is to be combined with prior information is presented. Application of the method in the case where the objective is to estimate volume per acre and prior knowledge is represented by a yield equation demonstrates that this method can reduce the amount of sample information that would be required if the yield equation were to be ignored.

DETECTION OF TRACKS IN AERIAL PHOTOS BY THE GIBBS SAMPLER

2003 ◽  
Vol 17 (01) ◽  
pp. 1-16 ◽  
Author(s):  
A. TETERUKOVSKIY
Keyword(s):  
Prior Knowledge ◽  
Cell Size ◽  
Bayesian Approach ◽  
Gibbs Sampler ◽  
Prior Information ◽  
A Priori ◽  
A Priori Knowledge ◽  
Gray Level ◽  
Aerial Photos ◽  
Positive Results

A problem of automatic detection of tracks in aerial photos is considered. We adopt a Bayesian approach and base our inference on an a priori knowledge of the structure of tracks. The probability of a pixel to belong to a track depends on how the pixel gray level differs from the gray levels of pixels in the neighborhood and on additional prior information. Several suggestions on how to formalize the prior knowledge about the shape of the tracks are made. The Gibbs sampler is used to construct the most probable configuration of tracks in the area. The method is applied to aerial photos with cell size of 1 sq. m. Even for detection of trails of width comparable with or smaller than the cell size, positive results can be achieved.

scholarly journals Multiview Self-Supervised Segmentation for OARs Delineation in Radiotherapy

2021 ◽  
Vol 2021 ◽  
pp. 1-5
Author(s):  
Cong Liu ◽  
Xiaofei Zhang ◽  
Wen Si ◽  
Xinye Ni
Keyword(s):  
State Of The Art ◽  
Organs At Risk ◽  
Representation Learning ◽  
Supervised Segmentation ◽  
Training Samples ◽  
Latent Space ◽  
Unlabelled Data ◽  
Quantitative Performance ◽  
Automatic Drawing ◽  
Negative Training

Radiotherapy has become a common treatment option for head and neck (H&N) cancer, and organs at risk (OARs) need to be delineated to implement a high conformal dose distribution. Manual drawing of OARs is time consuming and inaccurate, so automatic drawing based on deep learning models has been proposed to accurately delineate the OARs. However, state-of-the-art performance usually requires a decent amount of delineation, but collecting pixel-level manual delineations is labor intensive and may not be necessary for representation learning. Encouraged by the recent progress in self-supervised learning, this study proposes and evaluates a novel multiview contrastive representation learning to boost the models from unlabelled data. The proposed learning architecture leverages three views of CTs (coronal, sagittal, and transverse plane) to collect positive and negative training samples. Specifically, a CT in 3D is first projected into three 2D views (coronal, sagittal, and transverse planes), then a convolutional neural network takes 3 views as inputs and outputs three individual representations in latent space, and finally, a contrastive loss is used to pull representation of different views of the same image closer (“positive pairs”) and push representations of views from different images (“negative pairs”) apart. To evaluate performance, we collected 220 CT images in H&N cancer patients. The experiment demonstrates that our method significantly improves quantitative performance over the state-of-the-art (from 83% to 86% in absolute Dice scores). Thus, our method provides a powerful and principled means to deal with the label-scarce problem.

scholarly journals Semi-supervised segmentation and genome annotation

2020 ◽  
Author(s):  
Rachel C.W. Chan ◽  
Matthew McNeil ◽  
Eric G. Roberts ◽  
Mickaël Mendez ◽  
Maxwell W. Libbrecht ◽  
...  
Keyword(s):  
Supervised Learning ◽  
Prior Knowledge ◽  
Genome Annotation ◽  
Whole Genome ◽  
Transcription Start ◽  
Transcription Start Sites ◽  
Annotation Method ◽  
Supervised Segmentation ◽  
Unseen Data ◽  
Genome Annotations

AbstractSegmentation and genome annotation methods automatically discover joint signal patterns in whole genome datasets. Previously, researchers trained these algorithms in a fully unsupervised way, with no prior knowledge of the functions of particular regions. Adding information provided by expert-created annotations to supervise training could improve the annotations created by these methods. We implemented semi-supervised learning using virtual evidence in the annotation method Segway. Additionally, we defined a positionally tolerant precision and recall metric for scoring genome annotations based on the proximity of each annotation feature to the truth set. We demonstrate semi-supervised Segway’s ability to learn patterns corresponding to provided transcription start sites on a specified supervision label, and subsequently recover other transcription start sites in unseen data on the same supervision label.

scholarly journals Analysis of small-angle scattering data using model fitting and Bayesian regularization

2018 ◽  
Vol 51 (4) ◽  
pp. 1151-1161 ◽  
Author(s):  
Andreas Haahr Larsen ◽  
Lise Arleth ◽  
Steen Hansen
Keyword(s):  
Prior Knowledge ◽  
Small Angle ◽  
Regularization Method ◽  
Prior Information ◽  
Form Factors ◽  
Small Angle Scattering ◽  
Scattering Data ◽  
Model Parameters ◽  
Bayesian Regularization ◽  
Angle Scattering

The structure of macromolecules can be studied by small-angle scattering (SAS), but as this is an ill-posed problem, prior knowledge about the sample must be included in the analysis. Regularization methods are used for this purpose, as already implemented in indirect Fourier transformation and bead-modeling-based analysis of SAS data, but not yet in the analysis of SAS data with analytical form factors. To fill this gap, a Bayesian regularization method was implemented, where the prior information was quantified as probability distributions for the model parameters and included via a functional S. The quantity Q = χ2 + αS was then minimized and the value of the regularization parameter α determined by probability maximization. The method was tested on small-angle X-ray scattering data from a sample of nanodiscs and a sample of micelles. The parameters refined with the Bayesian regularization method were closer to the prior values as compared with conventional χ2 minimization. Moreover, the errors on the refined parameters were generally smaller, owing to the inclusion of prior information. The Bayesian method stabilized the refined values of the fitted model upon addition of noise and can thus be used to retrieve information from data with low signal-to-noise ratio without risk of overfitting. Finally, the method provides a measure for the information content in data, N g, which represents the effective number of retrievable parameters, taking into account the imposed prior knowledge as well as the noise level in data.

scholarly journals How do the Questions Asked Affect Suspects’ Perceptions of the Interviewer’s Prior Knowledge?

2019 ◽  
Author(s):  
Meghana Srivatsav ◽  
Timothy John Luke ◽  
Pär Anders Granhag ◽  
Aldert Vrij
Keyword(s):  
Prior Knowledge ◽  
Prior Information ◽  
Relevance Theory ◽  
Related Topic ◽  
Underlying Mechanisms ◽  
Interview Transcript ◽  
Different Levels

The aim of this study was to understand if guilty suspects’ perceptions regarding the prior information or evidence held by the interviewer against the suspect could be influenced through the content of the investigative questions. To test this idea, we explored three question-phrasing factors that we labeled as Topic Discussion (if a specific crime-related topic was discussed or not), Specificity (different levels of crime-related details included in the questions) and Stressor (emphasis on the importance of the specific crime-related detail in the questions). The three factors were chosen based on relevance theory, a psycholinguistic theory that explores how people draw inferences from the communicated content. Participants (N= 370) assumed the role of the suspect and read a crime narrative and an interview transcript based on the suspect’s activities. After reading the narrative and the transcripts, participants responded to scales that measured their perception of interviewer’s prior knowledge (PIK) regarding the suspects’ role in the crime, based on the questions posed by the interviewer in the transcripts. Of the three factors tested, we found that questioning about a specific crime-related topic (Topic Discussion) increased their PIK. This study is the first to explore the underlying mechanisms of how suspects draw inferences regarding the interviewer’s prior knowledge through the content of the investigative questions adopting concepts of psycholinguistic theory.

Sign in / Sign up

Export Citation Format

Share Document