Mean Intersection Over Union Research Articles

A highly efficient tunnel lining crack detection model based on Mini-Unet.

The accurate detection of tunnel lining cracks and prompt identification of their primary causes are critical for maintaining tunnel availability. The advancement of deep learning, particularly in the domain of convolutional neural network (CNN) for image segmentation, has made tunnel lining crack detection more feasible. However, the CNN-based technique for tunnel lining crack detection commonly prioritizes increasing algorithmic complexity to enhance detection accuracy, posing a challenge in balancing the accuracy of detection and the efficiency of the algorithm. Motivated by the superior performance of Unet in image segmentation, this paper proposes a lightweight tunnel lining crack detection model named Mini-Unet, which refined the Unet architecture and utilized depthwise separable convolutions (DSConv) to replace some standard convolution layers. In the optimization of the proposed model parameters, applying a hybrid loss function that integrated dice loss and cross-entropy loss effectively tackled the imbalance between crack and background categories. Several models were set up to contrast with Mini-Unet and the experimental results were analyzed. Mini-Unet achieves a mean intersection over union (MIoU) of 60.76%, a mean precision of 84.18%, and a frame per second (FPS) of 5.635, respectively. Mini-Unet outperforms several mainstream models, enabling rapid detection while maintaining identified accuracy and facilitating the practical application of AI power for real-time tunnel lining crack detection.

TIA-UNet: transformer-enhanced deep learning for adolescent idiopathic scoliosis spinal x-ray image segmentation

Abstract Adolescent Idiopathic Scoliosis (AIS) is a common spinal deformity where precise diagnosis is crucial for developing effective treatment strategies. Traditional manual x-ray image analysis is time-consuming and highly dependent on the operator’s expertise, thus constraining diagnostic efficiency and accuracy. This study aimed to develop an automated thoracolumbar spine segmentation method utilizing deep learning to enhance the efficiency and accuracy of AIS diagnosis. We introduced TIA-UNet, an innovative network architecture that combines Convolutional Neural Networks (CNN) with Transformer models. By integrating the IR Block and DA Block, TIA-UNet was optimized for feature extraction and multi-scale information fusion. The model underwent training and validation using our established Adolescent Scoliosis Medical Dataset (ASMD). TIA-UNet attained a Dice similarity coefficient of 90. 02%, a Mean Intersection over Union (MIoU) of 81. 96%, and a Hausdorff Distance (HD) of 4. 09, significantly surpassing current state-of-the-art methods such as UNet and TransUNet. Moreover, TIA-UNet exhibited superior computational efficiency regarding parameter count, inference time, and floating-point operations (FLOPs). As an automated medical image segmentation algorithm, TIA-UNet enhanced segmentation accuracy while maintaining high computational efficiency, demonstrating significant potential for clinical diagnostic applications. This study provides compelling evidence supporting the utilization of deep learning techniques in medical image analysis.

Enhancing detection of various pancreatic lesions on endoscopic ultrasound through artificial intelligence: a basis for computer-aided detection systems.

Endoscopic ultrasound (EUS) is the most sensitive method for evaluation of pancreatic lesions but is limited by significant operator dependency. Artificial intelligence (AI), in the form of computer-aided detection (CADe) systems, has shown potential in increasing accuracy and bridging operator dependency in several endoscopic domains. However, the complexity of integrating AI into EUS is far more challenging. This aims to develop and test the basis for a CADe system for real-time detection and segmentation of all pancreatic lesions. In this single-center study EUS studies of pancreatic findings were included. Lesions were outlined by two expert (>5years performing EUS) endoscopists, and the two leading types of models were benchmarked. The models' performance was evaluated through per-pixel intersection over union (IoU). A total of 1497 EUS images from 165 patients were evaluated. The dataset included malignancies, neuroendocrine tumors, benign cysts, chronic and acute pancreatitis, normal fatty pancreas, and benign lesions. The best model demonstrated detection and segmentation on the test set with a mean IoU of 0.73, achieving a PPV, NPV, total accuracy, and ROC of 0.82, 0.96, 0.95, and 0.95, respectively. The algorithm is adaptable for real-time processing. We developed and tested deep learning models for real-time detection and segmentation of pancreatic lesions during EUS with promising results. This constitutes the basis for a CADe system for EUS, which could be valuable in future detection and evaluation of pancreatic lesions. Further studies for validation and generalization are underway.

Automatic reconstruction of semantic façade model of architectural heritage

AbstractFaçade elements, such as windows, doors, balconies, sculptures, and totems, in architectural heritage images with incomplete structures should be automatically reconstructed for applications in 3D analysis, 3D modeling, virtual tourism, city planning, and the protection and reconstruction of architectural heritage. This study segments façade elements of architectural heritage semantically using YOLOv9. A parameterized expression for the semantic façade model is designed. In addition, the façade layer graph (FLG) and element layer graph (ELG) algorithms are developed based on topological, geometric, and structural constraints to automatically reconstruct the semantic façade model for architectural heritages. The results showed that the average precision (AP) and mean intersection over union (MIoU) achieved using YOLOv9 + FLG-ELG are 86.91% and 85.63%, respectively, on the dataset concerning façade elements of architectural heritages. The AP values obtained from the proposed method are 98.5% on the ECP2011 dataset and 95.3% on the Graz2012 dataset. The YOLOv9 + FLG-ELG method automatically reconstructs regular, irregular, and complex façade layouts with high accuracy and robustness.

The Segment Anything foundation model achieves favorable brain tumor auto-segmentation accuracy in MRI to support radiotherapy treatment planning.

Promptable foundation auto-segmentation models like Segment Anything (SA, Meta AI, New York, USA) represent anovel class of universal deep learning auto-segmentation models that could be employed for interactive tumor auto-contouring in RT treatment planning. Segment Anything was evaluated in an interactive point-to-mask auto-segmentation task for glioma brain tumor auto-contouring in 16,744 transverse slices from 369 MRI datasets (BraTS 2020 dataset). Up to nine interactive point prompts were automatically placed per slice. Tumor boundaries were auto-segmented on contrast-enhanced T1w sequences. Out of the three auto-contours predicted by SA, accuracy was evaluated for the contour with the highest calculated IoU (Intersection over Union, "oracle mask," simulating interactive model use with selection of the best tumor contour) and for the tumor contour with the highest model confidence ("suggested mask"). Mean best IoU (mbIoU) using the best predicted tumor contour (oracle mask) in full MRI slices was 0.762 (IQR 0.713-0.917). The best 2D mask was achieved after amean of 6.6 interactive point prompts (IQR 5-9). Segmentation accuracy was significantly better for high- compared to low-grade glioma cases (mbIoU 0.789 vs. 0.668). Accuracy was worse using the suggested mask (0.572). Stacking best tumor segmentations from transverse MRI slices, mean 3D Dice score for tumor auto-contouring was 0.872, which was improved to 0.919 by combining axial, sagittal, and coronal contours. The Segment Anything foundation segmentation model can achieve high accuracy for glioma brain tumor segmentation in MRI datasets. The results suggest that foundation segmentation models could facilitate RT treatment planning when properly integrated in aclinical application.

Recognition of palaeo-glacial troughs with cirques on parts of Tibetan Plateau using multi-modal datasets with deep learning models

Recognition of landforms associated with past glaciation is crucial for understanding past ice dynamics and their relationship to climate. With the development of artificial intelligence technology, deep learning techniques have assisted in the automatic extraction of glacial landforms, but these methods still face problems of low precision and weak transferability. This study proposes a new method named geomorphology-attention DeeplabV3+ (GA-DeeplabV3+) model. This method adds spatial attention and channel attention modules based on the DeeplabV3+ network and utilizes a combination of multi-modal geographic data. Verification shows that the model proposed in this paper significantly enhances the recognition precision of glacial troughs with cirques compared to existing models, reaching a Mean Intersection over Union (MIoU) of 86.2% and a mean Pixel Accuracy (mPA) of 90.64% in the area of the Palaeo-Daocheng ice cap. In addition, validation experiments were conducted in the Peiku Gangri region and the Tenasserim mountains, achieving MIoU scores of 70.09% and 73.28% respectively. This accomplishment represents a vital stride towards automating the extraction of palaeo-glacial landforms, which holds great significance for analyzing the scale and evolution of ancient glaciers.

Research on Traffic Marking Segmentation Detection Algorithm Based on Feature Fusion

Identifying road markings is a very important part of the vehicle environment sensing system and plays a crucial role in a vehicle’s correct understanding of a current traffic situation. However, road traffic markings are interfered with by a variety of factors, such as being obscured and the viewpoint of the vehicle sensors, resulting in large errors in the existing detection methods. In order to make the target detection task applicable to irregular objects or to detection tasks with higher accuracy requirements while reducing the waste of computational resources, this paper improves the accuracy of traffic marking segmentation detection by designing a multi-type traffic marking segmentation detection model based on image segmentation algorithms and designing a segmentation guidance matrix module based on a rank guidance matrix computation method. By constructing a comprehensive traffic marking detection model, a unified road traffic marking detection is achieved. Finally, the new traffic marking datasets ApolloScape-Precise and ApolloScape-Large are constructed based on the existing ApolloScape dataset, and experimental validation is carried out on these two datasets. The results show that the index MIoU (Mean Intersection over Union) of traffic marking segmentation detection reaches 61.44% and 70.15%, thus achieving a more perfect road traffic marking detection and right-of-way information perception and proving the effectiveness of the integrated traffic marking detection method designed in this paper.

Automated segmentation of retinal layers in optical coherence tomography images using Xception70 feature extraction

Optical coherence tomography (OCT) imaging plays a critical role in evaluating retinal layer thickness, serving as a pivotal diagnostic tool for numerous retinal conditions. However, challenges such as speckle noise, poor image contrast, and ambiguous retinal detachments, like drusen, often hinder accurate segmentation of retinal layers. To address these challenges and enhance diagnostic precision, we introduce the Retinal Segmentation Network, or "Ret-Seg Net," a deep neural network-based approach. Leveraging the advanced Xception70 feature extractor, Ret-Seg Net extracts and comprehends the intricate properties of retinal layers. By integrating acquired feature maps from Xception70 into the atrous spatial pyramid-pooling module, Ret-Seg Net extracts multiscale feature information. The encoder-decoder module of Ret-Seg Net achieves automated segmentation of retinal layers in OCT images by reconstructing distinct retinal layer borders. This advanced module accurately recognizes and differentiates retinal layer boundaries, providing precise and reliable segmentation. Validation of our approach using real-time images and the Duke dataset, comprising 310 volumes with 40 B-scans each, demonstrates outstanding performance. Mean intersection over union (MIoU) and sensitivity (Se) metrics achieved remarkable values of 94.52 % and 96.25 % respectively. Furthermore, our approach offers a versatile segmentation framework applicable to various tissues and cell types in clinical settings. Automating segmentation of retinal layers enhances precision in disease identification and monitoring while significantly improving labor efficiency by reducing the need for manual segmentation.

Hybrid Feature Extraction for Breast Cancer Classification Using the Ensemble Residual VGG16 Deep Learning Model

Introduction: Breast Cancer (BC) is a significant cause of high mortality amongst women globally and probably will remain a disease posing challenges about its detectability. Advancements in medical imaging technology have improved the accuracy and efficiency of breast cancer classification. However, tumor features' complexity and imaging data variability still pose challenges. Method: This study proposes the Ensemble Residual-VGG-16 model as a novel combination of the Deep Residual Network (DRN) and VGG-16 architecture. This model is purposely engineered with maximal precision for the task of breast cancer diagnosis based on mammography images. We assessed its performance by accuracy, recall, precision, and the F1-Score. All these metrics indicated the high performance of this Residual-VGG-16 model. The diagnostic residual-VGG16 performed exceptionally well with an accuracy of 99.6%, precision of 99.4%, recall of 99.7%, F1 score of 98.6%, and Mean Intersection over Union (MIoU) of 99.8% with MIAS datasets. Result: Similarly, the INBreast dataset achieved an accuracy of 93.8%, a precision of 94.2%, a recall of 94.5%, and an F1-score of 93.4% Conclusion: The proposed model is a significant advancement in breast cancer diagnosis, with high accuracy and potential as an automated grading.

A semi‐supervised segmentation network fusing pseudo‐label with multi‐level feature consistency correction for hard exudates

AbstractTimely detection of hard exudates in fundus images can effectively avoid the severity of the disease, but the labelling of small and numerous lesion areas requires a lot of labour costs. This paper proposes a semi‐supervised segmentation network, which integrates pseudo‐labels and multi‐level features consistency correction. It achieves accurate segmentation of hard exudates by making full use of a small amount of labelled data and a large amount of unlabelled data. The network effectively extracts features from the unlabelled data through knowledge transfer of the teacher‐student model, and incorporates a Transformer network for auxiliary training to promote the quality of transfer. In addition, three unsupervised losses are introduced to improve the performance: the perturbation loss improves the robustness of the model to noise by adding different noises to the same input; the multi‐level feature consistency correction loss ensures the consistency of features of the student model at different scales; and the pseudo‐labelling cross‐supervision loss utilizes the generated pseudo‐labels for supervision between CNN and Transformer. By comparing the segmentation results with different proportion of the labelled data, it has better segmentation performance compared to other methods. The proposed methods can totally increase dice by 16.56% and mean intersection over union (MIoU) by 25.11%.

FG-UNet: fine-grained feature-guided UNet for segmentation of weeds and crops in UAV images.

Semantic segmentation of weed and crop images is a key component and prerequisite for automated weed management. For weeds in unmanned aerial vehicle (UAV) images, which are usually characterized by small size and easily confused with crops at early growth stages, existing semantic segmentation models have difficulties to extract sufficiently fine features. This leads to their limited performance in weed and crop segmentation of UAV images. We proposed a fine-grained feature-guided UNet, named FG-UNet, for weed and crop segmentation in UAV images. Specifically, there are two branches in FG-UNet, namely the fine-grained feature branch and the UNet branch. In the fine-grained feature branch, a fine feature-aware (FFA) module was designed to mine fine features in order to enhance the model's ability to segment small objects. In the UNet branch, we used an encoder-decoder structure to realize high-level semantic feature extraction in images. In addition, a contextual feature fusion (CFF) module was designed for the fusion of the fine features and high-level semantic features, thus enhancing the feature discrimination capability of the model. The experimental results showed that our proposed FG-UNet, achieved state-of-the-art performance compared to other semantic segmentation models, with mean intersection over union (MIOU) and mean pixel accuracy (MPA) of 88.06% and 92.37%, respectively. The proposed method in this study lays a solid foundation for accurate detection and intelligent management of weeds. It will have a positive impact on the development of smart agriculture. © 2024 Society of Chemical Industry.

HSP-UNet: An Accuracy and Efficient Segmentation Method for Carbon Traces of Surface Discharge in the Oil-Immersed Transformer

Restricted by a metal-enclosed structure, the internal defects of large transformers are difficult to visually detect. In this paper, a micro-robot is used to visually inspect the interior of a transformer. For the micro-robot to successfully detect the discharge level and insulation degradation trend in the transformer, it is essential to segment the carbon trace accurately and rapidly from the complex background. However, the complex edge features and significant size differences of carbon traces pose a serious challenge for accurate segmentation. To this end, we propose the Hadamard production-Spatial coordinate attention-PixelShuffle UNet (HSP-UNet), an innovative architecture specifically designed for carbon trace segmentation. To address the pixel over-concentration and weak contrast of carbon trace image, the Adaptive Histogram Equalization (AHE) algorithm is used for image enhancement. To realize the effective fusion of carbon trace features with different scales and reduce model complexity, the novel grouped Hadamard Product Attention (HPA) module is designed to replace the original convolution module of the UNet. Meanwhile, to improve the activation intensity and segmentation completeness of carbon traces, the Spatial Coordinate Attention (SCA) mechanism is designed to replace the original jump connection. Furthermore, the PixelShuffle up-sampling module is used to improve the parsing ability of complex boundaries. Compared with UNet, UNet++, UNeXt, MALUNet, and EGE-UNet, HSP-UNet outperformed all the state-of-the-art methods on both carbon trace datasets. For dendritic carbon traces, HSP-UNet improved the Mean Intersection over Union (MIoU), Pixel Accuracy (PA), and Class Pixel Accuracy (CPA) of the benchmark UNet by 2.13, 1.24, and 4.68 percentage points, respectively. For clustered carbon traces, HSP-UNet improved MIoU, PA, and CPA by 0.98, 0.65, and 0.83 percentage points, respectively. At the same time, the validation results showed that the HSP-UNet has a good model lightweighting advantage, with the number of parameters and GFLOPs of 0.061 M and 0.066, respectively. This study could contribute to the accurate segmentation of discharge carbon traces and the assessment of the insulation condition of the oil-immersed transformer.

USCT-UNet: Rethinking the Semantic Gap in U-Net Network from U-shaped Skip Connections with Multichannel Fusion Transformer.

Medical image segmentation is a crucial component of computer-aided clinical diagnosis, with state-of-the-art models often being variants of U-Net. Despite their success, these models' skip connections introduce an unnecessary semantic gap between the encoder and decoder, which hinders their ability to achieve the high precision required for clinical applications. Awareness of this semantic gap and its detrimental influences have increased over time. However, a quantitative understanding of how this semantic gap compromises accuracy and reliability remains lacking, emphasizing the need for effective mitigation strategies. In response, we present the first quantitative evaluation of the semantic gap between corresponding layers of U-Net and identify two key characteristics: 1) The direct skip connection (DSC) exhibits a semantic gap that negatively impacts models' performance; 2) The magnitude of the semantic gap varies across different layers. Based on these findings, we re-examine this issue through the lens of skip connections. We introduce a Multichannel Fusion Transformer (MCFT) and propose a novel USCT-UNet architecture, which incorporates U-shaped skip connections (USC) to replace DSC, allocates varying numbers of MCFT blocks based on the semantic gap magnitude at different layers, and employs a spatial channel cross-attention (SCCA) module to facilitate the fusion of features between the decoder and USC. We evaluate USCT-UNet on four challenging datasets, and the results demonstrate that it effectively eliminates the semantic gap. Compared to using DSC, our USC and SCCA strategies achieve maximum improvements of 4.79% in the Dice coefficient, 5.70% in mean intersection over union (MIoU), and 3.26 in Hausdorff distance.

A globally sampled high-resolution hand-labeled validation dataset for evaluating surface water extent maps

Abstract. Effective monitoring of global water resources is increasingly critical due to climate change and population growth. Advancements in remote sensing technology, specifically in spatial, spectral, and temporal resolutions, are revolutionizing water resource monitoring, leading to more frequent and high-quality surface water extent maps using various techniques such as traditional image processing and machine learning algorithms. However, satellite imagery datasets contain trade-offs that result in inconsistencies in performance, such as disparities in measurement principles between optical (e.g., Sentinel-2) and radar (e.g., Sentinel-1) sensors and differences in spatial and spectral resolutions among optical sensors. Therefore, developing accurate and robust surface water mapping solutions requires independent validations from multiple datasets to identify potential biases within the imagery and algorithms. However, high-quality validation datasets are expensive to build, and few contain information on water resources. For this purpose, we introduce a globally sampled, high-spatial-resolution dataset labeled using 3 m PlanetScope imagery (Planet Team, 2017). Our surface water extent dataset comprises 100 images, each with a size of 1024×1024 pixels, which were sampled using a stratified random sampling strategy covering all 14 biomes. We highlighted urban and rural regions, lakes, and rivers, including braided rivers and coastal regions. We evaluated two surface water extent mapping methods using our dataset – Dynamic World (Brown et al., 2022), based on Sentinel-2, and the NASA IMPACT model (Paul and Ganju, 2021), based on Sentinel-1. Dynamic World achieved a mean intersection over union (IoU) of 72.16 % and F1 score of 79.70 %, while the NASA IMPACT model had a mean IoU of 57.61 % and F1 score of 65.79 %. Performance varied substantially across biomes, highlighting the importance of evaluating models on diverse landscapes to assess their generalizability and robustness. Our dataset can be used to analyze satellite products and methods, providing insights into their advantages and drawbacks. Our dataset offers a unique tool for analyzing satellite products, aiding the development of more accurate and robust surface water monitoring solutions. The dataset can be accessed via https://doi.org/10.25739/03nt-4f29 (Mukherjee et al., 2024).

End-to-End Fully Automated Lung Cancer Volume Estimation System

The volume of the tumor plays a very crucial role in deciding the stage of lung cancer which in turn helps in deciding the best treatment and its schedule. Currently used computer-based volume estimation techniques are semi-automatic with limited accuracy. For any automatic lung cancer segmentation system, lung CT scans of hundreds of patients are required along with their corresponding annotated segmentation masks. It is difficult to get accurately annotated data as cancer segmentation of CT scans done by the radiologists, is a time-consuming manual process. Also, it is subjective and prone to intra and inter-observer variability. Further, owing to the irregular shape of the cancerous tumor, accurate volume estimation becomes a challenge with regular convolution models. This paper proposes an end-to-end automatic tumor volume estimation model that estimates volume using the GPR (Gaussian Process Regression) interpolation method. The proposed modified cancer segmentation model uses deformable convolutions. This modification offers a higher segmentation accuracy in terms of IoU (Intersection over Union) and clearly defined nodule boundaries with correct retention of the nodule shape. The research was undertaken in collaboration with Nanavati Hospital, Mumbai, and all the models were validated on a real dataset obtained from the hospital. The proposed model gives a mean segmentation IoU (Intersection over Union) of 0.9035 and a volume estimation accuracy of 93.13% which are almost 5% and 3% higher than 0.8548 and 90.51% which are the corresponding results obtained using a standard U-net++ algorithm.

Electrical Impedance Tomography-Based Electronic Skin for Multi-Touch Tactile Sensing Using Hydrogel Material and FISTA Algorithm.

Flexible electronic skin (e-skin) can enable robots to have sensory forms similar to human skin, enhancing their ability to obtain more information from touch. The non-invasive nature of electrical impedance tomography (EIT) technology allows electrodes to be arranged only at the edges of the skin, ensuring the stretchability and elasticity of the skin's interior. However, the image quality reconstructed by EIT technology has deteriorated in multi-touch identification, where it is challenging to clearly reflect the number of touchpoints and accurately size the touch areas. This paper proposed an EIT-based flexible tactile sensor that employs self-made hydrogel material as the primary sensing medium. The sensor's structure, fabrication process, and tactile imaging principle were elaborated. To improve the quality of image reconstruction, the fast iterative shrinkage-thresholding algorithm (FISTA) was embedded into the EIDORS toolkit. The performances of the e-skin in aspects of assessing the touching area, quantitative force sensing and multi-touch identification were examined. Results showed that the mean intersection over union (MIoU) of the reconstructed images was improved up to 0.84, and the tactile position can be accurately imaged in the case of the number of the touchpoints up to seven (larger than two to four touchpoints in existing studies), proving that the combination of the proposed sensor and imaging algorithm has high sensitivity and accuracy in multi-touch tactile sensing. The presented e-skin shows potential promise for the application in complex human-robot interaction (HRI) environments, such as prosthetics and wearable devices.

Intelligent classification of water bodies with different turbidity levels based on Gaofen-1 multispectral imagery

The turbidity of water is crucial for the health of river and lake ecosystems, necessitating efficient monitoring for effective water management. Existing methods for studying water turbidity's spatial and temporal distribution rely mostly on measured data. There is limited research on the classification of water bodies with different turbidity levels. The main challenge lies in determining the boundaries of liquid water bodies at various turbidity levels, making it challenging to classify them accurately using traditional remote sensing image classification methods. This paper proposes and validates an intelligent turbidity classification method based on deep learning using GaoFen-1 multispectral remote sensing imagery. An adaptive threshold water extraction method based on the Normalized Difference Water Index is proposed to capture water boundaries more accurately to improve the accuracy of extracting nearshore water bodies. A semi-automatic semantic annotation method for water turbidity is introduced to reduce manual labeling costs. The paper applies mode filtering to address edge noise issues and establishes a high-quality training sample dataset. After comparing the accuracy of various neural network models, DeepLab V3+ is selected for intelligent turbidity classification. The results show high accuracy, with mean intersection over union (MIoU), mean F1 score (MF1), and overall accuracy (OA) reaching 94.73%, 97.29%, and 97.54%, respectively. The proposed method and experiments demonstrate the feasibility of intelligent classification of water bodies with different turbidity levels using deep learning networks. This provides a new approach for large-scale and efficient remote sensing water turbidity monitoring.

Lightweight multi-scale feature dense cascade neural network for scene understanding of intelligent autonomous platform

Understanding the surrounding environment is an essential part of intelligent autonomous platform to complete autonomous driving or other autonomous tasks. Thereinto, scene understanding is a key technology to understand environmental information. For intelligent autonomous platform, it not only requires high accuracy, but also has a strict requirement for the inference speed. Moreover, it should be able to perform normally in complex environments such as target occlusion, the influence of illumination, and objects at different sizes. For tackling these problems, in this study, a lightweight multi-scale feature dense cascade neural network (LMFDCNet) is proposed in real-time scene understanding in complex scenes. Furthermore, trilateral fusion module is adopted to enhance low-level detailed features or high-level semantic features of each branch to describe the object better in different environments. A multi-scale upsampling module that fuses contextual information at different scales allows objects to be represented at different size. The results indicate that LMFDCNet produces 99.1 % classification accuracy on the tobacco dataset. On the Cityscapes dataset, we achieve 74.2 % Mean Intersection over Union (MIoU) with 13.6 M parameters at the speed of 38.2 FPS on a single 1070Ti card. To test the performance on the real-time application in the real world, LMFDCNet is deployed on the autonomous platform for real-time vision task such as classification of tobacco leaf state in curing process and semantic segmentation in the driving process of autonomous platform. The results reveal that LMFDCNet perform well on the AI embedded device and could be applied into intelligent autonomous platform.

Lightweight deep neural network for radio frequency interference detection and segmentation in synthetic aperture radar

Radio frequency interference (RFI) poses challenges in the analysis of synthetic aperture radar (SAR) images. Existing RFI suppression systems rely on prior knowledge of the presence of RFI. This paper proposes a lightweight neural network-based algorithm for detecting and segmenting RFI (LDNet) in the time-frequency domain. The network accurately delineates RFI pixel regions in time-frequency spectrograms. To mitigate the impact on the operational speed of the entire RFI suppression system, lightweight modules and pruning operations are introduced. Compared to threshold-based RFI detection algorithms, deep learning-based segmentation networks, and AC-UNet specifically designed for RFI detection, LDNet achieves improvements in mean intersection over union (MIoU) by 24.56%, 13.29%, and 7.54%, respectively.Furthermore, LDNet reduces model size by 99.03% and inference latency by 24.53% compared to AC-UNet.

An Optimized Semi-Supervised Generative Adversarial Network Rice Extraction Method Based on Time-Series Sentinel Images

In response to the limitations of meteorological conditions in global rice growing areas and the high cost of annotating samples, this paper combines the Vertical-Vertical (VV) polarization and Vertical-Horizontal (VH) polarization backscatter features extracted from Sentinel-1 synthetic aperture radar (SAR) images and the NDVI, NDWI, and NDSI spectral index features extracted from Sentinel-2 multispectral images. By leveraging the advantages of an optimized Semi-Supervised Generative Adversarial Network (optimized SSGAN) in combining supervised learning and semi-supervised learning, rice extraction can be achieved with fewer annotated image samples. Within the optimized SSGAN framework, we introduce a focal-adversarial loss function to enhance the learning process for challenging samples; the generator module employs the Deeplabv3+ architecture, utilizing a Wide-ResNet network as its backbone while incorporating dropout layers and dilated convolutions to improve the receptive field and operational efficiency. Experimental results indicate that the optimized SSGAN, particularly when utilizing a 3/4 labeled sample ratio, significantly improves rice extraction accuracy, leading to a 5.39% increase in Mean Intersection over Union (MIoU) and a 2.05% increase in Overall Accuracy (OA) compared to the highest accuracy achieved before optimization. Moreover, the integration of SAR and multispectral data results in an OA of 93.29% and an MIoU of 82.10%, surpassing the performance of single-source data. These findings provide valuable insights for the extraction of rice information in global rice-growing regions.