A Novel Method for Recognizing Space Radiation Sources Based on Multi-Scale Residual Prototype Learning Network

Deep residual networks (ResNets) can learn deep feature representation from hyperspectral images (HSIs), and therefore have been widely used for HSI classification. Despite their high accuracies, there still exist a lot of challenging cases, such as open world recognition, limited-sample learning and visualization of learned classification features, which cannot be well addressed. Most of the challenges in HSI classification can be attributed to the dependence on softmax based loss function and classifiers, which cause the lack of robustness for deep learning models and the hardness to visualize the learned classification features. To improve the robustness and achieve the visualization of learned classification features, we propose a novel learning framework called Residual Prototype Learning Network, a combination of residual network and prototypes learning mechanism. Under the framework, a prototype learning based loss function is proposed to enhance intra-class compactness and the inter-class separation of these feature representations; in addition, a prototype learning based classifier is simultaneously proposed to achieve the 2D or 3D visualization of the classification features. The effectiveness of our proposed learning framework is evaluated on several publicly available HSI benchmarks, and the experimental results show that our approach achieve better results than traditional softmax based ResNets.

In recent years, CNN has been used for single image super-resolution (SR) with its success of in the field of computer vision. However, in the recovery process, there are always some high-frequency components that cant be recovered from low-resolution images to high-resolution ones by using existing CNN-based methods. In this paper, we propose an image super-resolution method based on CNN, which uses a two-level residual learning network to learn residual components, i.e., high-frequency components. We use the Super-Resolution Convolutional Neural Network (SRCNN) as the network structure in each level so that our proposed method can achieve the high-resolution images with high-frequency components that cant be obtained by the existing methods. In addition, we analyze the proposed method with considering three kinds of residual learning networks, which are different in the structure and superimposed layers of the residual learning network. In the experiments, we investigate the performance of the proposed method with various residual learning networks and the effect of image super-resolution to image captioning task.

The minimum distance between adjacent shots is always “larger” than that of adjacent receivers in each shot, and the inconsistency can decrease the performance of migration. Interpolation methods can improve the data consistency, while most methods are suitable for randomly missing cases, and the interpolation difficulty increases sharply for regularly missing cases, especially for the cases with a big gap. As deep learning has a strong self-learning ability in nonlinear characterization, we introduce it into shot gather reconstruction. First, the residual learning networks (ResNets) with better back propagation property is illustrated, and the self-similarity of the common shot and receiver gathers are analyzed theoretically followed by briefly discussing the interpolation issue. Then, the common shot gathers are divided into the training sets to train ResNets and the validation sets to verify the trained ResNets. Finally, the trained ResNets is used to reconstruct the missing shot gathers in the common receiver gather. Three different datasets are used to demonstrate the reconstruction validity of the proposed deep learning strategy. The reconstructed data with better consistency can improve the accuracy of the final reservoir characterization. Presentation Date: Tuesday, October 16, 2018 Start Time: 1:50:00 PM Location: 204B (Anaheim Convention Center) Presentation Type: Oral

With the emergence of text-to-speech (TTS) systems, the synthetic speech have been widely applied. But the quality evaluation method of the synthetic speech is still away from an agreement. This paper presents a novel method based on residual learning networks, which was trained on the databases of Blizzard Challenge and our own corpus. The effect of the proposed architecture was compared with the LSTM networks and VGG networks. According to the results of experiments, the architecture based on the ResNet gives root mean square error (RMSE) of 0.40 and the correlation coefficient of 0.82, which is better than the other two networks.

Marine mammal vocalizations provide a reliable method of identifying most species. These vocalizations may be effectively captured by the passive acoustic monitoring of underwater acoustic signals, but this approach generates prohibitively large amounts of data, making a machine learning approach desirable. In this study, we demonstrate that residual learning networks may be highly effective at classifying vocalizations by species and investigate the utility of multi-channel spectrograms in training these networks. We find that classification performance can be improved by effectively preprocessing the acoustic data. A series of tests are used to assess optimal parameters for noise-removal, spectrographic window functions, preprocessing augmentations, and multi-channel spectrogram generation. While we find that our networks do learn additional information from multi-channel spectrograms, we demonstrate that single-channel spectrograms still offer superior classification performance in 32-class classification tasks.

Current image super-resolution (SR) reconstruction methods based on deep learning focus more on using the learning capabilities of deeper networks to obtain better reconstruction results, but the increase in depth will increase the complexity of the network structure and the difficulty of training. To address this problem, considering the advantages of residual learning that can overcome the problem of low learning efficiency, and the ability of multiscale networks that can extract global and local features at different scales, this paper proposes a multi-scale residual block (MSRB) for feature learning. Based on the constructed MSRB, a cascaded multi-scale residual network (CMSRN) is developed for image SR reconstruction. In the network, to reconstruct richer image texture details, multiple multi-scale residual blocks are cascaded to construct the residual feature learning part. Experimental results on four datasets show that the proposed network can obtain better reconstruction results, and is superior to state-of-the-art SR reconstruction methods in terms of subjective observation and objective quantitative evaluation.

Due to the favorable denoising performance of the discriminative learning model, a large number of discriminative learning models have been designed to remove noise. However, most discriminative learning models can only deal with the noise that already exists in the training data. In this paper, a multi-scale trainable deep residual convolutional neural network (DCMSNet) based on dilated convolution is proposed. DCMSNet consists of a chain of dilated convolution layers, convolution layers, normalized multi-scale convolution blocks (BNMCBlock), multi-scale convolution blocks (MCBlock) and multi-scale dropout convolution blocks (MCDBlock). The use of dilated convolution can avoid over-parameterization problems caused by too deep networks. In order to capture more feature information during feature extraction and image reconstruction, we designed novel BNMCBlock and MCBlock. In order to reduce the degree of coupling between image features and improve the generalization ability of the network, we also designed a MCDBlock. Meanwhile, residual learning, batch normalization and dropout are utilized to speed up the training process and boost the denoising performance. Unlike existing denoising models, DCMSNet is able to remove different degrees of noise. Compared with state-of-the-art image denoising methods, DCMSNet has achieved relatively competitive results.

Breast cancer, which attacks the glandular epithelium of the breast, is the second most common kind of cancer in women after lung cancer, and it affects a significant number of people worldwide. Based on the advantages of Residual Convolutional Network and the Transformer Encoder with Multiple Layer Perceptron (MLP), this study proposes a novel hybrid deep learning Computer-Aided Diagnosis (CAD) system for breast lesions. While the backbone residual deep learning network is employed to create the deep features, the transformer is utilized to classify breast cancer according to the self-attention mechanism. The proposed CAD system has the capability to recognize breast cancer in two scenarios: Scenario A (Binary classification) and Scenario B (Multi-classification). Data collection and preprocessing, patch image creation and splitting, and artificial intelligence-based breast lesion identification are all components of the execution framework that are applied consistently across both cases. The effectiveness of the proposed AI model is compared against three separate deep learning models: a custom CNN, the VGG16, and the ResNet50. Two datasets, CBIS-DDSM and DDSM, are utilized to construct and test the proposed CAD system. Five-fold cross validation of the test data is used to evaluate the accuracy of the performance results. The suggested hybrid CAD system achieves encouraging evaluation results, with overall accuracies of 100% and 95.80% for binary and multiclass prediction challenges, respectively. The experimental results reveal that the proposed hybrid AI model could identify benign and malignant breast tissues significantly, which is important for radiologists to recommend further investigation of abnormal mammograms and provide the optimal treatment plan.

Synthetic aperture radar (SAR) image super-resolution (SR) has broad application prospects in military reconnaissance, marine monitoring, crop monitoring, and other fields. However, when the scale factor is large, current SAR image SR has limitations, such as insufficient feature extraction, difficulty of feature representation, and modeling difficulty of SR reconstruction. These problems have severely restricted the SR performance of SAR images. In addition, optical remote sensing images share higher resolution and richer high-frequency information than SAR remote sensing images. Therefore, we adopt high-resolution (HR) optical images as auxiliary images of SAR image SR to improve the performance of SAR image SR. Our research is conducted from two aspects: the extraction efficiency of HR optical image-guided information and the representation ability of low-resolution (LR) SAR image features. We propose an optical guidance residual networks (OGRNs) based on residual learning for SAR image SR. The OGRN uses optical images during network training, which enhances the performance of SAR image SR, even when optical images are missing during network testing. First, we extract the texture feature of SAR images through the designed residual learning module. Second, we extract the weight map of feature information for SAR image SR from HR optical images by the convolutional layer. Finally, the weight map is used to guide the weighted supervised SR reconstruction for SAR images. Extensive experimental results demonstrated that, under the guidance of optical images, OGRN achieves excellent performance in terms of both quantitative assessment metrics and visual quality.

Medical imaging plays an essential role in modern healthcare, providing non-invasive tools for diagnosing and monitoring various medical conditions. However, the resolution limitations of imaging hardware often result in suboptimal images, which can hinder the precision of clinical decision-making. Single image super-resolution (SISR) techniques offer a solution by reconstructing high-resolution (HR) images from low-resolution (LR) counterparts, enhancing the visual quality of medical images. In this paper, we propose an enhanced Residual Feature Learning Network (RFLN) tailored specifically for medical imaging. Our contributions include replacing the residual local feature blocks with standard residual blocks, increasing the model depth for improved feature extraction, and incorporating enhanced spatial attention (ESA) mechanisms to refine the feature selection. Extensive experiments on medical imaging datasets demonstrate that the proposed model achieves superior performance in terms of both quantitative metrics, such as PSNR and SSIM, and qualitative visual quality compared to existing state-of-the-art models. The enhanced RFLN not only effectively mitigates noise but also preserves critical anatomical details, making it a promising solution for high-precision medical imaging applications.

3D Point clouds are irregular and unordered. Recently, neural networks applying on point clouds have shown superior performance on 3D object classification and segmentation. PointNet has achieved competitive performance in classification tasks, however, it ignores the important local information of point clouds, which carries the important information in 3D geometric space. The informations lack will reduce the accuracy of object classification. In this paper, a new deep neural network model is proposed. Specifically, a multiscale network is used to extract local features of points to improve the accuracy of PointNet and a deep residual network is used to avoid the vanishing gradient problem in the process of deep training network. Experiments show that the proposed network architecture achieves better performance than state-of-the-art methods on multiple challenging benchmark datasets and tasks.

In this paper, we propose a new deep network architecture named deep boosting denoising net (DBDnet) for image denoising. It is a residual learning network that can generate a noise map from a noisy observation. In detail, it first generates a coarse noise map via a simple structure, and then updates the noise map gradually via a boosting function. The motivation of our DBDnet stems from the observation that the noise map recovered by any algorithm cannot ideally equal the ground-truth noise map, which typically contains noise. We call this noise NoN, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e.</i> , noise of noise map. Based on this observation, we formulate the denoising as a process of reducing NoN, and the role of DBDnet is to eliminate the NoN from the coarse noise map. In particular, we analyze the process of reducing NoN theoretically, and propose an NoN eliminating module to simulate it accordingly. We evaluate the proposed DBDnet on images polluted by different levels of additive white Gaussian noise and real noise. Experiment results demonstrate that our DBDnet can attain better denoising performance compared with state-of-the-art methods on several kinds of image denoising tasks. In particular, for the Gaussian denoising and real image denoising tasks, the average improvements of the PSNR values brought by our DBDnet are about 0.25 dB and 1.01 dB, respectively. In addition, we find and verify that the deep boosting insight can be easily introduced into the state-of-the-art image denoising network, and promotes its denoising performance. Our code is publicly available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/jiayi-ma/DBDNet</uri> .

Human activity recognition by radar sensors plays an important role in healthcare and smart homes. However, labeling a large number of radar datasets is difficult and time-consuming, and it is difficult for models trained on insufficient labeled data to obtain exact classification results. In this paper, we propose a multiscale residual weighted classification network with large-scale, medium-scale, and small-scale residual networks. Firstly, an MRW image encoder is used to extract salient feature representations from all time-Doppler images through contrastive learning. This can extract the representative vector of each image and also obtain the pre-training parameters of the MRW image encoder. During the pre-training process, large-scale residual networks, medium-scale residual networks, and small-scale residual networks are used to extract global information, texture information, and semantic information, respectively. Moreover, the time-channel weighting mechanism can allocate weights to important time and channel dimensions to achieve more effective extraction of feature information. The model parameters obtained from pre-training are frozen, and the classifier is added to the backend. Finally, the classifier is fine-tuned using a small amount of labeled data. In addition, we constructed a new dataset with eight dangerous activities. The proposed MRW-CN model was trained on this dataset and achieved a classification accuracy of 96.9%. We demonstrated that our method achieves state-of-the-art performance. The ablation analysis also demonstrated the role of multi-scale convolutional kernels and time-channel weighting mechanisms in classification.

Recurrent reverse attention guided residual learning for saliency object detection

With the rapid growth of Internet video image information, there is a large amount of redundancy in image data. Use less data stream information to transfer the image or the amount of information contained in the image. Its purpose is to reduce the redundancy of images, so as to store them at low bit rate and reduce the data storage space. In the general image compression method, the hybrid coding framework is adopted. Each algorithm adopts a fixed algorithm mode through a specific design algorithm, without global optimization. Image compression is mainly divided into prediction, transformation, quantization, digital entropy coding and other steps. At present, there are many researches on super-resolution network based on deep learning technology. The main function is to reconstruct high-resolution image replace image magnification low-resolution images such as linear interpolation, which has a great performance improvement image resolution, noise reduction, deblurring and so on, but there is no effective way to use super-resolution network applications to improve quality of compression reconstructed image quality. This paper involves a new method that using image super-resolution residual learning network to improve quality of compression image, our method, the reduced image is encoded into a content stream and a transmission corresponding parameter is encoded into a model stream. Firstly, the original image is scaled down 1/2 size of source image, then encode the small image into content stream with the existing codec. Secondly, the residual learning super-resolution (SR) network is used for image filtering to scale up reconstructed image with decode image resizing method and boost the quality of edge feature extraction of image. Our results show that there is significant performance improvement of h265 in low resolution reconstructed image (bits-per-pixel less than 0.1).