Deforestation is a major concern for preserving the biodiversity of the entire globe. During the last few years, machine learning and deep learning methods have been employed for mapping deforestation. There is still ...
详细信息
Deforestation is a major concern for preserving the biodiversity of the entire globe. During the last few years, machine learning and deep learning methods have been employed for mapping deforestation. There is still scope for ample improvement in these methods as they are prone to errors and can give inaccurate results because of over or under-segmentation. This paper uses deep convolutional neural network-based semantic segmentation to process multispectral satellite images to monitor forest cover changes in Tadoba-Andhari National Park during the period 2000-2022. The proposed approach uses the U-Net architecture with extended inputs which gives more accuracy as compared to U-Net with only image input. Landsat images along with vegetation indices have been used as training data. The proposed method requires less time to train the model and is also cost-efficient in terms of computing requirements. The performance of the proposed method was compared with state-of-the-art methods where the proposed method outperformed the other models with an F1-score of 0.90 and an accuracy of 84.83%. When compared with U-Net trained with Landsat images only, it was observed that the U-Net model trained with extended input was able to achieve better results.
In this paper, we present a new edge detection model based on proximal unfolded neural networks. The architecture relies on unfolding proximal Blake-Zisserman iterations, leading to a composition of two blocks: a smoo...
详细信息
In this paper, we present a new edge detection model based on proximal unfolded neural networks. The architecture relies on unfolding proximal Blake-Zisserman iterations, leading to a composition of two blocks: a smoothing block and an edge detection block. We show through simulations that the proposed approach efficiently eliminates irrelevant details while retaining key edges and significantly improves performance with respect to state-of-the-art strategies. Additionally, our architecture is significantly lighter than recent learning models designed for edge detection in terms of number of learnable parameters and inference time.
Bearings are vital components in rotating machinery. Undetected bearing faults may result not only in financial loss, but also in the loss of lives. Hence, there exists an abundance of studies working on the early det...
详细信息
Bearings are vital components in rotating machinery. Undetected bearing faults may result not only in financial loss, but also in the loss of lives. Hence, there exists an abundance of studies working on the early detection of bearing faults. The rising use of deep learning in recent years increased the number of imaging types/neural network architectures used for bearing fault classification, making it challenging to choose the most suitable 2-D imaging method and neural network. This study aims to address this challenge, by sharing the results of the training of eighteen imaging methods with four different networks using the same vibration data and training metrics. To further strengthen the results, the validation dataset size was taken as five times the training dataset size. The best results obtained is 99.89% accuracy by using Scattergram Filter Bank 1 as the image input, and ResNet-50 as the network for training. Prior to our work, Scattergram images have never been used for bearing fault classification. Ten out of 72 methods used in this work resulted in accuracies higher than 99.5%.
With neural networks growing deeper and feature maps growing larger, limited communication bandwidth with external memory (or DRAM) and power constraints become a bottleneck in implementing network inference on mobile...
详细信息
ISBN:
(纸本)9798350344868;9798350344851
With neural networks growing deeper and feature maps growing larger, limited communication bandwidth with external memory (or DRAM) and power constraints become a bottleneck in implementing network inference on mobile and edge devices. In this paper, we propose an end-to-end differentiable bandwidth efficient neural inference method with the activation compressed by neural data compression method. Specifically, we propose a transform-quantization-entropy coding pipeline for activation compression with symmetric exponential Golomb coding and a data-dependent Gaussian entropy model for arithmetic coding. Optimized with existing model quantization methods, low-level task of image compression can achieve up to 19x bandwidth reduction with 6.21x energy saving. The code implementation is available at https://***/xyzysz/Bandwidth_efficient_nic.
In this paper, we propose an innovative image enhancement algorithm called "Dual-Enhancing-Dense-UNet (DEDUNet)" that simultaneously performs image brightness enhancement and reduces noise. This model is bas...
详细信息
In this paper, we propose an innovative image enhancement algorithm called "Dual-Enhancing-Dense-UNet (DEDUNet)" that simultaneously performs image brightness enhancement and reduces noise. This model is based on Convolutional neural Network (CNN) algorithms and incorporates innovative techniques such as Decoupled Fully Connection (DFC) attention, skip connections, shortcut, Cross-Stage-Partial (CSP) and dense blocks to address the brightness enhancement and noise removal aspects of image enhancement. The dual approach to image enhancement offers a new solution for restoring and improving high-quality images, presenting new opportunities in the fields of computer vision and imageprocessing. Our experimental results substantiate the superior performance of the proposed algorithm, showcasing significant improvements in key performance indicators. Specifically, the algorithm achieves a Peak signal-to-Noise Ratio (PSNR) of 19.17, Structural Similarity Index (SSIM) of 0.71, Learned Perceptual image Patch Similarity (LPIPS) of 0.30, Mean Absolute Error (MAE) of 0.09, and a Multiply-Accumulate (MAC) of 0.696G. These results highlight the algorithm's remarkable image quality enhancement capabilities, demonstrating a considerable advantage over existing methods. Experimental results demonstrate the superior performance and efficiency of the proposed algorithm in terms of image quality improvement compared to existing methods.
Recent single image super-resolution methods based on various complex deep neural networks have achieved remarkable success. However, these methods require a large amount of computational overhead while improving perf...
详细信息
Recent single image super-resolution methods based on various complex deep neural networks have achieved remarkable success. However, these methods require a large amount of computational overhead while improving performance, and thus are difficult to apply to mobile devices in real-world scenarios. In this letter, we design an efficient asymmetric convolutional distillation block (ACDB). Especially in this block, introducing an asymmetric convolution block (ACB) and reusing shallow distillation features can effectively improve the performance of the model and reduce the model complexity. Our model achieves efficient performance while maintaining low complexity.
The performance of visual processing is commonly constrained in extreme outside weather such as heavy rain. Rain streaks may substantially damage image optical quality and impact imageprocessing in many scenarios. Th...
详细信息
The performance of visual processing is commonly constrained in extreme outside weather such as heavy rain. Rain streaks may substantially damage image optical quality and impact imageprocessing in many scenarios. Thus, it has practical application value in researching the problem of single image rain removal. However, removing rain streaks from a single image is a challenging task. Although end-to-end learning approaches based on convolutional neural networks have lately made significant progress on this task, most existing methods still cannot perform deraining well. They fail to process the details of the background layer, resulting in the loss of certain information. To address this issue, we propose a single image deraining network named twin-stage Unet-like network (TUNet). Specifically, a reconstitution residual block (RRB) is presented as the basic structure of encoder-decoder to obtain more spatial contextual information for extracting rain components. Then, a residual sampling module (RSM) is introduced to perform downsampling and upsampling operations to preserve residual properties in the structure while obtaining deeper image features. Finally, the convolutional block attention module (CBAM) is adopted to fuse shallow and deep features of the same size in the model. Extensive experiments on five publicly synthetic datasets and a real-world dataset demonstrate that our proposed TUNet model outperforms the state-of-the-art deraining approaches. The average PSNR value of TUNet is 0.41 dB higher than the state-of-the-art method (OSAM-Net) on synthetic datasets.
The COVID-19 virus is increasingly crucial to human health since new variants appear frequently. Detection of COVID-19 through respiratory sound has been an important area of research. This study analyzes respiratory ...
详细信息
The COVID-19 virus is increasingly crucial to human health since new variants appear frequently. Detection of COVID-19 through respiratory sound has been an important area of research. This study analyzes respiratory sounds using novel accumulated bi-spectral features. The principal domain bispectrum is used for computing accumulated bispectrum. The resulting magnitude bispectrum is used in forming the bispectral image. In this work, a convolutional neural network (CNN) and ResNet-50 algorithms are designed to classify respiratory sounds as either COVID-19 or healthy. The performance of the proposed method is compared with the state-of-the-art methods. The proposed CNN-based method achieves the highest accuracy of 87.68% for shallow breath sounds, and ResNet-50 achieves the highest accuracy of 87.62% for deep breath sounds. Similarly, proposed methods gives the improved performance for other respiratory sounds.
This paper addresses two key limitations in existing imagesignalprocessing (ISP) approaches: the suboptimal performance in low-light conditions and the lack of trainability in traditional ISP methods. To tackle thes...
详细信息
ISBN:
(纸本)9798350344868;9798350344851
This paper addresses two key limitations in existing imagesignalprocessing (ISP) approaches: the suboptimal performance in low-light conditions and the lack of trainability in traditional ISP methods. To tackle these issues, we propose a novel, trainable ISP framework that incorporates both the strengths of traditional ISP techniques and advanced MultiScale Retinex (MSR) algorithms for night-time enhancement. Our method consists of three primary components: an ISP-based Luminance Harmonization layer to initially optimize luminance levels in RAW data, a deep learning-based MSR layer for nuanced decomposition of image components, and a specialized enhancement layer for both precise, regionspecific luminance enhancement and color denoising. The proposed approach is validated through rigorous experiments on machine vision benchmarks and objective visual quality indicators. Our results demonstrate not only a significant improvement over existing methods but also robust adaptability under diverse lighting conditions. This work offers a versatile ISP framework with promising applications beyond its immediate scope.
Underwater image recognition plays a crucial role in assessing the health status of marine ecosystems. By utilizing underwater cameras and image recognition technology, researchers can monitor the biodiversity, popula...
详细信息
Underwater image recognition plays a crucial role in assessing the health status of marine ecosystems. By utilizing underwater cameras and image recognition technology, researchers can monitor the biodiversity, population numbers, growth status, and overall structure and functionality of ecosystems in the ocean. However, the problem of marine ecology assessment always occurs in dynamic and open environments, and discoveries of unknown new species are often made. Existing works which applied classification methods directly may not address this situation well. Therefore, unsupervised learning is needed to cluster these newly emerged species. However, due to strong noise interference in underwater images, clustering the unlabeled samples directly is difficult. To address this issue, we propose a two-stage training framework that can learn discriminative knowledge from labeled data for clustering new classes. Its core idea is to utilize pseudo-labeling to train the model, and then strengthens the capability of clustering by leveraging the consistency between the labeled and unlabeled data. Furthermore, contrastive learning is also used to optimize the model's representation in the embedding space. Experimental results on the WildFish dataset of over 5000 species verified the effectiveness of the proposed method in open-set underwater image recognition.
暂无评论