Self-Matching CAM: A Novel Accurate Visual Explanation of CNNs for SAR Image Interpretation

Feng, Zhenpeng; Zhu, Mingzhe; Stanković, Ljubiša; Ji, Hongbing

doi:10.3390/rs13091772

Cited by 39 publications

(13 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our method, we use the result of LIME to guide the selection of SAR images. In addition, we can carry out some feature engineering work based on the result of LIME, such as removing some misleading features to focus on the generation of SAR target [21]. At a deeper level, the result of LIME can guide us to select the best model [12], i.e., we can create a model that automatically uses the visualization results to guide the training of networks.…”

Section: Discussionmentioning

confidence: 99%

“…Propagation-based methods visualize the correlation between input and output by decomposing the result of the output layer and propagating it to the input space layer by layer according to the designed propagation rules [16,17]. CAM methods visualize the area of interest of the model by providing a highlighted region to reflect the networks' interest in a specific class by weighted summation of feature maps [18][19][20][21][22]. Perturbationbased methods observe the change of the output by masking, deleting, or blurring different regions of the input, thereby, determining the impact of the corresponding region on the output [23,24].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

LIME-Based Data Selection Method for SAR Images Generation Using GAN

et al. 2022

Self Cite

View full text Add to dashboard Cite

Deep learning has obtained remarkable achievements in computer vision, especially image and video processing. However, in synthetic aperture radar (SAR) image recognition, the application of DNNs is usually restricted due to data insufficiency. To augment datasets, generative adversarial networks (GANs) are usually used to generate numerous photo-realistic SAR images. Although there are many pixel-level metrics to measure GAN’s performance from the quality of generated SAR images, there are few measurements to evaluate whether the generated SAR images include the most representative features of the target. In this case, the classifier probably categorizes a SAR image into the corresponding class based on “wrong” criterion, i.e., “Clever Hans”. In this paper, local interpretable model-agnostic explanation (LIME) is innovatively utilized to evaluate whether a generated SAR image possessed the most representative features of a specific kind of target. Firstly, LIME is used to visualize positive contributions of the input SAR image to the correct prediction of the classifier. Subsequently, these representative SAR images can be selected handily by evaluating how much the positive contribution region matches the target. Experimental results demonstrate that the proposed method can ally “Clever Hans” phenomenon greatly caused by the spurious relationship between generated SAR images and the corresponding classes.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

LIME-Based Data Selection Method for SAR Images Generation Using GAN

et al. 2022

Self Cite

View full text Add to dashboard Cite

show abstract

“…In recent years, heatmap-based visualization methods that unravel the internal mechanisms of deep learning models have been developed. Although there are a number of proposed methodologies, the class activation mapping (CAM) methods (e.g., GRAD-CAM) are becoming increasingly popular [120].…”

Section: Nirs-based Imaging Applications For the Detection Of Chemica...mentioning

confidence: 99%

Near-Infrared Hyperspectral Imaging Pipelines for Pasture Seed Quality Evaluation: An Overview

Reddy

Guthridge

Panozzo

et al. 2022

Sensors

View full text Add to dashboard Cite

Near-infrared (800–2500 nm; NIR) spectroscopy coupled to hyperspectral imaging (NIR-HSI) has greatly enhanced its capability and thus widened its application and use across various industries. This non-destructive technique that is sensitive to both physical and chemical attributes of virtually any material can be used for both qualitative and quantitative analyses. This review describes the advancement of NIR to NIR-HSI in agricultural applications with a focus on seed quality features for agronomically important seeds. NIR-HSI seed phenotyping, describing sample sizes used for building high-accuracy calibration and prediction models for full or selected wavelengths of the NIR region, is explored. The molecular interpretation of absorbance bands in the NIR region is difficult; hence, this review offers important NIR absorbance band assignments that have been reported in literature. Opportunities for NIR-HSI seed phenotyping in forage grass seed are described and a step-by-step data-acquisition and analysis pipeline for the determination of seed quality in perennial ryegrass seeds is also presented.

show abstract

“…Shanshan Shang et al [22] proposed bidimensional intrinsic mode functions (BIMFs) for SAR ATR, which is the combination of multi-mode representations extracted by bidimensional empirical mode decomposition (BEMD) and ResNet. Zhenpeng Feng et al [23] proposed a Self-Matching class activation mapping (CAM) to improve the interpretability of SAR images.…”

Section: Introductionmentioning

confidence: 99%

A Lightweight Fully Convolutional Neural Network for SAR Automatic Target Recognition

Zhou

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

Automatic target recognition (ATR) in synthetic aperture radar (SAR) images has been widely used in civilian and military fields. Traditional model-based methods and template matching methods do not work well under extended operating conditions (EOCs), such as depression angle variant, configuration variant, and noise corruption. To improve the recognition performance, methods based on convolutional neural networks (CNN) have been introduced to solve such problems and have shown outstanding performance. However, most of these methods rely on continuously increasing the width and depth of networks. This adds a large number of parameters and computational overhead, which is not conducive to deployment on edge devices. To solve these problems, a novel lightweight fully convolutional neural network based on Channel-Attention mechanism, Channel-Shuffle mechanism, and Inverted-Residual block, namely the ASIR-Net, is proposed in this paper. Specifically, we deploy Inverted-Residual blocks to extract features in high-dimensional space with fewer parameters and design a Channel-Attention mechanism to distribute different weights to different channels. Then, in order to increase the exchange of information between channels, we introduce the Channel-Shuffle mechanism into the Inverted-Residual block. Finally, to alleviate the matter of the scarcity of SAR images and strengthen the generalization performance of the network, four approaches of data augmentation are proposed. The effect and generalization performance of the proposed ASIR-Net have been proved by a lot of experiments under both SOC and EOCs on the MSTAR dataset. The experimental results indicate that ASIR-Net achieves higher recognition accuracy rates under both SOC and EOCs, which is better than the existing excellent ATR methods.

show abstract

Self-Matching CAM: A Novel Accurate Visual Explanation of CNNs for SAR Image Interpretation

Cited by 39 publications

References 26 publications

LIME-Based Data Selection Method for SAR Images Generation Using GAN

LIME-Based Data Selection Method for SAR Images Generation Using GAN

Near-Infrared Hyperspectral Imaging Pipelines for Pasture Seed Quality Evaluation: An Overview

A Lightweight Fully Convolutional Neural Network for SAR Automatic Target Recognition

Contact Info

Product

Resources

About