Comparison of Backbones for Semantic Segmentation Network

Zhang, Rongyu; Du, Lixuan; Qu, Xiao; Liu, Jiaming

doi:10.1088/1742-6596/1544/1/012196

Cited by 39 publications

(19 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this study, only U-Net was used among the CNN architectures, but regional performance comparisons with other CNN models are essential for the normalization regional performance differences [32]. We will conduct investigations to quantify and examine regional performance differences according to the architecture by using more advanced architectures such as the ResNet [33], GoogLeNet [34], and AlexNet [35] series in the future.…”

Section: Discussionmentioning

confidence: 99%

Segmentation Performance Comparison Considering Regional Characteristics in Chest X-ray Using Deep Learning

Lee

Kim

2022

Sensors

View full text Add to dashboard Cite

Chest radiography is one of the most widely used diagnostic methods in hospitals, but it is difficult to read clearly because several human organ tissues and bones overlap. Therefore, various image processing and rib segmentation methods have been proposed to focus on the desired target. However, it is challenging to segment ribs elaborately using deep learning because they cannot reflect the characteristics of each region. Identifying which region has specific characteristics vulnerable to deep learning is an essential indicator of developing segmentation methods in medical imaging. Therefore, it is necessary to compare the deep learning performance differences based on regional characteristics. This study compares the differences in deep learning performance based on the rib region to verify whether deep learning reflects the characteristics of each part and to demonstrate why this regional performance difference has occurred. We utilized 195 normal chest X-ray datasets with data augmentation for learning and 5-fold cross-validation. To compare segmentation performance, the rib image was divided vertically and horizontally based on the spine, clavicle, heart, and lower organs, which are characteristic indicators of the baseline chest X-ray. Resultingly, we found that the deep learning model showed a 6–7% difference in the segmentation performance depending on the regional characteristics of the rib. We verified that the performance differences in each region cannot be ignored. This study will enable a more precise segmentation of the ribs and the development of practical deep learning algorithms.

show abstract

Section: Discussionmentioning

confidence: 99%

Segmentation Performance Comparison Considering Regional Characteristics in Chest X-ray Using Deep Learning

Lee

Kim

2022

Sensors

View full text Add to dashboard Cite

show abstract

“…Additionally, botanists showed interest to apply up-to-date DL in HTPP data analysis ( Jiang and Li, 2020 ). The encoder and decoder portions of U-Net ( Figure 1E ) showed several performances in various environments ( Zhang et al, 2020 ). The encoder provides valuable information on whether various encoders at U-Net yield different results for interesting traits (PA) in agriculture ( Jiang and Li, 2020 ).…”

Section: Discussionmentioning

confidence: 99%

“…The second half of the architecture, the decoder, uses features from the previous step. For separating object and background information, advanced encoders gather additional features from images and achieve higher accuracies ( Hoeser and Kuenzer, 2020 ; Zhang et al, 2020 ). Hence, for segmenting, there is room for improvement because U-Net performs well in different soil conditions.…”

Section: Introductionmentioning

confidence: 99%

Time-Series Growth Prediction Model Based on U-Net and Machine Learning in Arabidopsis

et al. 2021

View full text Add to dashboard Cite

Yield prediction for crops is essential information for food security. A high-throughput phenotyping platform (HTPP) generates the data of the complete life cycle of a plant. However, the data are rarely used for yield prediction because of the lack of quality image analysis methods, yield data associated with HTPP, and the time-series analysis method for yield prediction. To overcome limitations, this study employed multiple deep learning (DL) networks to extract high-quality HTTP data, establish an association between HTTP data and the yield performance of crops, and select essential time intervals using machine learning (ML). The images of Arabidopsis were taken 12 times under environmentally controlled HTPP over 23 days after sowing (DAS). First, the features from images were extracted using DL network U-Net with SE-ResXt101 encoder and divided into early (15–21 DAS) and late (∼21–23 DAS) pre-flowering developmental stages using the physiological characteristics of the Arabidopsis plant. Second, the late pre-flowering stage at 23 DAS can be predicted using the ML algorithm XGBoost, based only on a portion of the early pre-flowering stage (17–21 DAS). This was confirmed using an additional biological experiment (P < 0.01). Finally, the projected area (PA) was estimated into fresh weight (FW), and the correlation coefficient between FW and predicted FW was calculated as 0.85. This was the first study that analyzed time-series data to predict the FW of related but different developmental stages and predict the PA. The results of this study were informative and enabled the understanding of the FW of Arabidopsis or yield of leafy plants and total biomass consumed in vertical farming. Moreover, this study highlighted the reduction of time-series data for examining interesting traits and future application of time-series analysis in various HTPPs.

show abstract

“…In this paper, we will compare the advantages and disadvantages of various classical basic classification networks as backbones through experiments [23].…”

Section: Methodsmentioning

confidence: 99%

Comparison of Semantic Convolution Neural Networks on the Example of Crack Segmentation in Asphalt Images

Akimov¹,

Мустафина²,

Plotnikova³

2021

IJC

View full text Add to dashboard Cite

The article is devoted to a comparative analysis of the effectiveness of convolutional neural networks for semantic segmentation of road surface damage marking. Currently, photo and video surveillance methods are used to control the condition of the road surface. Assessing and analyzing new manual data can take too long. Thus, a completely different procedure is required to inspect and assess the state of controlled objects using technical vision. The authors compared 3 neural networks (Unet, Linknet, PSPNet) used in semantic segmentation using the example of the Crack500 dataset. The proposed architectures have been implemented in the Keras and TensorFlow frameworks. The compared models of neural convolutional networks effectively solve the assigned tasks even with a limited amount of training data. High accuracy is observed. The considered models can be used in various segmentation tasks. The results obtained can be used in the process of modeling, monitoring, and predicting the wear of the road surface.

show abstract

Comparison of Backbones for Semantic Segmentation Network

Cited by 39 publications

References 5 publications

Segmentation Performance Comparison Considering Regional Characteristics in Chest X-ray Using Deep Learning

Segmentation Performance Comparison Considering Regional Characteristics in Chest X-ray Using Deep Learning

Time-Series Growth Prediction Model Based on U-Net and Machine Learning in Arabidopsis

Comparison of Semantic Convolution Neural Networks on the Example of Crack Segmentation in Asphalt Images

Contact Info

Product

Resources

About