MFCSNet: Multi-Scale Deep Features Fusion and Cost-Sensitive Loss Function Based Segmentation Network for Remote Sensing Images

Wang, Ende; Jiang, Yanmei; Li, Yong; Jin, Yang; Ren, Mengcheng; Zhang, Qingchun

doi:10.3390/app9194043

Cited by 11 publications

(10 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the MSFFU-Net model proposed in this paper has better classification performance and the ability to detect more tiny vessels. It proves that the feature fusion decoder structure applied max-pooling indices can recorde more accurately the retinal vascular edge and location information, and the multi-scale feature extraction encoder based on Inception module can make the thin tiny retinal blood vessel features more discriminative, which can present excellent segmentation performance [ 35 ]. Therefore, it also demonstrates that the proposed model has better performance on retinal blood vessels segmentation than the U-Net model.…”

Section: Resultsmentioning

confidence: 99%

“…The cross-entropy loss function is defined as follows:

where

and

are the ground truth and prediction, respectively

is 0 or 1 and

is between 0 and 1. The cost-sensitive matrix

is incorporated, as shown in formula (3), this can avoid under-fitting due to the small number of retinal blood vessels during the neural network learning process [ 35 ]. When retinal blood vessels are misclassified, the cost will be greater and the attention to retinal blood vessels will be increased:

…”

Section: Proposed Methodsmentioning

confidence: 99%

See 1 more Smart Citation

A Multi-Scale Feature Fusion Method Based on U-Net for Retinal Vessel Segmentation

Yang

Liu

Ren

et al. 2020

Entropy

Self Cite

View full text Add to dashboard Cite

Computer-aided automatic segmentation of retinal blood vessels plays an important role in the diagnosis of diseases such as diabetes, glaucoma, and macular degeneration. In this paper, we propose a multi-scale feature fusion retinal vessel segmentation model based on U-Net, named MSFFU-Net. The model introduces the inception structure into the multi-scale feature extraction encoder part, and the max-pooling index is applied during the upsampling process in the feature fusion decoder of an improved network. The skip layer connection is used to transfer each set of feature maps generated on the encoder path to the corresponding feature maps on the decoder path. Moreover, a cost-sensitive loss function based on the Dice coefficient and cross-entropy is designed. Four transformations—rotating, mirroring, shifting and cropping—are used as data augmentation strategies, and the CLAHE algorithm is applied to image preprocessing. The proposed framework is tested and trained on DRIVE and STARE, and sensitivity (Sen), specificity (Spe), accuracy (Acc), and area under curve (AUC) are adopted as the evaluation metrics. Detailed comparisons with U-Net model, at last, it verifies the effectiveness and robustness of the proposed model. The Sen of 0.7762 and 0.7721, Spe of 0.9835 and 0.9885, Acc of 0.9694 and 0.9537 and AUC value of 0.9790 and 0.9680 were achieved on DRIVE and STARE databases, respectively. Results are also compared to other state-of-the-art methods, demonstrating that the performance of the proposed method is superior to that of other methods and showing its competitive results.

show abstract

Section: Resultsmentioning

confidence: 99%

“…The cross-entropy loss function is defined as follows:

where

and

are the ground truth and prediction, respectively

is 0 or 1 and

is between 0 and 1. The cost-sensitive matrix

…”

Section: Proposed Methodsmentioning

confidence: 99%

A Multi-Scale Feature Fusion Method Based on U-Net for Retinal Vessel Segmentation

Yang

Liu

Ren

et al. 2020

Entropy

Self Cite

View full text Add to dashboard Cite

show abstract

“…The decoder fused and evaluated image features by scale fusion and scale sensitivity. In 2019, Wang et al [77] designed multiscale trait detection methods based on encoding and decoding structure for fusing two different phases (such as low-level and high-level) denotative data.…”

Section: Edge Detection Technology Based On Encoding and Decodingmentioning

confidence: 99%

A Review of the Edge Detection Technology

Hou¹,

Jia²,

Wang³

et al. 2021

STAIQC

View full text Add to dashboard Cite

The edge detection-based has profoundly inspired recent works in image classification, object detection, segmentation, et al. With the growth of computer vision, the performance of edge detection has been notably improved. In this paper, we concentrate on presenting some edge detection technologies and grouping them into two major categories: classical edge detection technology and deep learning-based edge detection technology. For every team in the classification, the fundamental thoughts are first described, and then we probe beneath into the mainly related literature in recent years. Furthermore, the evaluation indicators are entirely elaborated. In addition, this conclusion is given by a comprehensive summary which provides suggestions on future perspectives.

show abstract

“…Two modules, channel feature compression (CFC) and multi-level feature aggregation upsample (MFAU), were designed to reduce the loss of details and make the edge clear. Moreover, Wang et al [28] defined a cost-sensitive loss function in addition to fuse multi-scale deep features.…”

Section: Introductionmentioning

confidence: 99%

Attentively Learning Edge Distributions for Semantic Segmentation of Remote Sensing Imagery

Chen

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

Semantic segmentation has been a fundamental task in interpreting remote sensing imagery (RSI) for various downstream applications. Due to the high intra-class variants and inter-class similarities, inflexibly transferring natural image-specific networks to RSI is inadvisable. To enhance the distinguishability of learnt representations, attention modules were developed and applied to RSI, resulting in satisfactory improvements. However, these designs capture contextual information by equally handling all the pixels regardless of whether they around edges. Therefore, blurry boundaries are generated, rising high uncertainties in classifying vast adjacent pixels. Hereby, we propose an edge distribution attention module (EDA) to highlight the edge distributions of leant feature maps in a self-attentive fashion. In this module, we first formulate and model column-wise and row-wise edge attention maps based on covariance matrix analysis. Furthermore, a hybrid attention module (HAM) that emphasizes the edge distributions and position-wise dependencies is devised combing with non-local block. Consequently, a conceptually end-to-end neural network, termed as EDENet, is proposed to integrate HAM hierarchically for the detailed strengthening of multi-level representations. EDENet implicitly learns representative and discriminative features, providing available and reasonable cues for dense prediction. The experimental results evaluated on ISPRS Vaihingen, Potsdam and DeepGlobe datasets show the efficacy and superiority to the state-of-the-art methods on overall accuracy (OA) and mean intersection over union (mIoU). In addition, the ablation study further validates the effects of EDA.

show abstract

MFCSNet: Multi-Scale Deep Features Fusion and Cost-Sensitive Loss Function Based Segmentation Network for Remote Sensing Images

Cited by 11 publications

References 31 publications

A Multi-Scale Feature Fusion Method Based on U-Net for Retinal Vessel Segmentation

A Multi-Scale Feature Fusion Method Based on U-Net for Retinal Vessel Segmentation

A Review of the Edge Detection Technology

Attentively Learning Edge Distributions for Semantic Segmentation of Remote Sensing Imagery

Contact Info

Product

Resources

About