Low-level Feature Extraction Research Articles

DDANet: A deep dilated attention network for intracerebral haemorrhage segmentation.

Intracranial haemorrhage (ICH) is an urgent and potentially fatal medical condition caused by brain blood vessel rupture, leading to blood accumulation in the brain tissue. Due to the pressure and damage it causes to brain tissue, ICH results in severe neurological impairment or even death. Recently, deep neural networks have been widely applied to enhance the speed and precision of ICH detection yet they are still challenged by small or subtle hemorrhages. The authors introduce DDANet, a novel haematoma segmentation model for brain CT images. Specifically, a dilated convolution pooling block is introduced in the intermediate layers of the encoder to enhance feature extraction capabilities of middle layers. Additionally, the authors incorporate a self-attention mechanism to capture global semantic information of high-level features to guide the extraction and processing of low-level features, thereby enhancing the model's understanding of the overall structure while maintaining details. DDANetalso integrates residual networks, channel attention, and spatial attention mechanisms for joint optimisation, effectively mitigating the severe class imbalance problem and aiding the training process. Experiments show that DDANet outperforms current methods, achieving the Dice coefficient, Jaccard index, sensitivity, accuracy, and a specificity of 0.712, 0.601, 0.73, 0.997, and 0.998, respectively. The code is available at https://github.com/hpguo1982/DDANet.

EfficientNet-resDDSC: A Hybrid Deep Learning Model Integrating Residual Blocks and Dilated Convolutions for Inferring Gene Causality in Single-Cell Data.

Gene Regulatory Networks (GRNs) reveal complex interactions between genes in organisms, crucial for understanding the life system's operation. The rapid development of biotechnology, especially single-cell RNA sequencing (scRNA-seq), has generated a large amount of scRNA-seq data, which can be analyzed to explore the regulatory relationships between genes at the single-cell level. Previous models used to construct GRNs mainly aim at constructing associative relationships between genes, but usually fail to accurately reveal the causality between genes. Therefore, we present a hybrid deep learning model called EfficientNet-resDDSC (the EfficientNet with Residual Blocks and Depthwise Separable Dilated Convolutions) to infer causality between genes. The model inherits the basic structure of EfficientNet-B0 and incorporates residual blocks as well as dilated convolutions. The model's ability to extract low-level features at the primary stage is enhanced by introducing residual blocks. The model combines Depthwise Separable Convolution (DSC) in the inverted linear bottleneck layers with the dilated convolutions to expand the model's receptive fields without increasing the computational effort. This design enables the model to comprehensively reveal potential relationships among different genes in high-dimensional and high-noise single-cell data. In comparison with the five existing deep learning network models, EfficientNet-resDDSC's overall performance is significantly better than others on four datasets. In this study, EfficientNet-resDDSC was further applied to construct GRNs for breast cancer patients, focusing on the related regulatory genes of the key gene BRCA1, which contributes to the advancement of breast cancer research and treatment strategies.

Multilevel hybrid handcrafted feature extraction based depression recognition method using speech

Background and purposeDiagnosis of depression is based on tests performed by psychiatrists and information provided by patients or their relatives. In the field of machine learning (ML), numerous models have been devised to detect depression automatically through the analysis of speech audio signals. While deep learning approaches often achieve superior classification accuracy, they are notably resource-intensive. This research introduces an innovative, multilevel hybrid feature extraction-based classification model, specifically designed for depression detection, which exhibits reduced time complexity. Materials and methodsMODMA dataset consisting of 29 healthy and 23 Major depressive disorder audio signals was used. The constructed model architecture integrates multilevel hybrid feature extraction, iterative feature selection, and classification processes. During the Hybrid Handcrafted Feature (HHF) generation stage, a combination of textural and statistical methods was employed to extract low-level features from speech audio signals. To enhance this process for high-level feature creation, a Multilevel Discrete Wavelet Transform (MDWT) was applied. This technique produced wavelet subbands, which were then input into the hybrid feature extractor, enabling the extraction of both high and low-level features. For the selection of the most pertinent features from these extracted vectors, Iterative Neighborhood Component Analysis (INCA) was utilized. Finally, in the classification phase, a one-dimensional nearest neighbor classifier, augmented with ten-fold cross-validation, was implemented to achieve detailed, results. ResultsThe HHF-based speech audio signal classification model attained excellent performance, with the 94.63 % classification accuracy. ConclusionsThe findings validate the remarkable proficiency of the introduced HHF-based model in depression classification, underscoring its computational efficiency.

Perceived duration of visual stimuli contracts due to crowding.

Recent research on duration perception suggests that duration encoding is not a single general process but involves several separate processes, some of which are specific to visual modality. Moreover, different functional aspects of visual processing can influence duration perception in distinct ways. One of the most important functions of the visual system is to identify and recognize features, shapes, and objects. However, it is still unclear whether and how computations related to these processes affect duration perception. To clarify this issue, we used a spatial crowding phenomenon, which allows the dissociation of low-level feature extraction from high-level processes such as object recognition. We created letter and vernier stimuli matched for their low-level properties but different in their discriminability due to spatial crowding. Here, we show that stimuli that became more difficult to discriminate appeared shorter in duration (data collected in 2019-2023). This difference in perceived duration could not be explained by low-level stimulus properties, cognitive bias due to discriminability, or perceived stimulus onsets or offsets. These results suggest the existence of time-sensitive structures specific to visual processing of features, shapes, and objects that is affected by crowding. These findings support the notion of distributed timing mechanisms in the visual system. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Image tampering detection based on RDS-YOLOv5 feature enhancement transformation

Malicious image tampering has gradually become another way to threaten social stability and personal safety. Timely detection and precise positioning can help reduce the occurrence of risks and improve the overall safety of society. Due to the limitations of highly targeted dataset training and low-level feature extraction efficiency, the generalization and actual performance of the recent tampered detection technology have not yet reached expectations. In this study, we propose a tampered image detection method based on RDS-YOLOv5 feature enhancement transformation. Firstly, a multi-channel feature enhancement fusion algorithm is proposed to enhance the tampering traces in tampered images. Then, an improved deep learning model named RDS-YOLOv5 is proposed for the recognition of tampered images, and a nonlinear loss metric of aspect ratio was introduced into the original SIOU loss function to better optimize the training process of the model. Finally, RDS-YOLOv5 is trained by combining the features of the original image and the enhancement image to improve the robustness of the detection model. A total of 6187 images containing three forms of tampering: splice, remove, and copy-move were used to comprehensively evaluate the proposed algorithm. In ablation test, compared with the original YOLOv5 model, RDS-YOLOv5 achieved a performance improvement of 6.46%, 5.13%, and 3.15% on F1-Score, mAP50 and mAP95, respectively. In comparative experiments, using SRIOU as the loss function significantly improved the model’s ability to search for the real tampered regions by 2.54%. And the RDS-YOLOv5 model trained by the fusion dataset further improved the comprehensive detection performance by about 1%.

GroupFormer for hyperspectral image classification through group attention

Hyperspectral image (HSI) data has a wide range of valuable spectral information for numerous tasks. HSI data encounters challenges such as small training samples, scarcity, and redundant information. Researchers have introduced various research works to address these challenges. Convolution Neural Network (CNN) has gained significant success in the field of HSI classification. CNN’s primary focus is to extract low-level features from HSI data, and it has a limited ability to detect long-range dependencies due to the confined filter size. In contrast, vision transformers exhibit great success in the HSI classification field due to the use of attention mechanisms to learn the long-range dependencies. As mentioned earlier, the primary issue with these models is that they require sufficient labeled training data. To address this challenge, we proposed a spectral-spatial feature extractor group attention transformer that consists of a multiscale feature extractor to extract low-level or shallow features. For high-level semantic feature extraction, we proposed a group attention mechanism. Our proposed model is evaluated using four publicly available HSI datasets, which are Indian Pines, Pavia University, Salinas, and the KSC dataset. Our proposed approach achieved the best classification results in terms of overall accuracy (OA), average accuracy (AA), and Kappa coefficient. As mentioned earlier, the proposed approach utilized only 5%, 1%, 1%, and 10% of the training samples from the publicly available four datasets.

Graph neural pre-training based drug-target affinity prediction.

Computational drug-target affinity prediction has the potential to accelerate drug discovery. Currently, pre-training models have achieved significant success in various fields due to their ability to train the model using vast amounts of unlabeled data. However, given the scarcity of drug-target interaction data, pre-training models can only be trained separately on drug and target data, resulting in features that are insufficient for drug-target affinity prediction. To address this issue, in this paper, we design a graph neural pre-training-based drug-target affinity prediction method (GNPDTA). This approach comprises three stages. In the first stage, two pre-training models are utilized to extract low-level features from drug atom graphs and target residue graphs, leveraging a large number of unlabeled training samples. In the second stage, two 2D convolutional neural networks are employed to combine the extracted drug atom features and target residue features into high-level representations of drugs and targets. Finally, in the third stage, a predictor is used to predict the drug-target affinity. This approach fully utilizes both unlabeled and labeled training samples, enhancing the effectiveness of pre-training models for drug-target affinity prediction. In our experiments, GNPDTA outperforms other deep learning methods, validating the efficacy of our approach.

Versatile Video Coding-Post Processing Feature Fusion: A Post-Processing Convolutional Neural Network with Progressive Feature Fusion for Efficient Video Enhancement

Advanced video codecs such as High Efficiency Video Coding/H.265 (HEVC) and Versatile Video Coding/H.266 (VVC) are vital for streaming high-quality online video content, as they compress and transmit data efficiently. However, these codecs can occasionally degrade video quality by adding undesirable artifacts such as blockiness, blurriness, and ringing, which can detract from the viewer’s experience. To ensure a seamless and engaging video experience, it is essential to remove these artifacts, which improves viewer comfort and engagement. In this paper, we propose a deep feature fusion based convolutional neural network (CNN) architecture (VVC-PPFF) for post-processing approach to further enhance the performance of VVC. The proposed network, VVC-PPFF, harnesses the power of CNNs to enhance decoded frames, significantly improving the coding efficiency of the state-of-the-art VVC video coding standard. By combining deep features from early and later convolution layers, the network learns to extract both low-level and high-level features, resulting in more generalized outputs that adapt to different quantization parameter (QP) values. The proposed VVC-PPFF network achieves outstanding performance, with Bjøntegaard Delta Rate (BD-Rate) improvements of 5.81% and 6.98% for luma components in random access (RA) and low-delay (LD) configurations, respectively, while also boosting peak signal-to-noise ratio (PSNR).

Research on Rail Surface Defect Detection Based on Improved CenterNet

Rail surface defect detection is vital for railway safety. Traditional methods falter with varying defect sizes and complex backgrounds, while two-stage deep learning models, though accurate, lack real-time capabilities. To overcome these challenges, we propose an enhanced one-stage detection model based on CenterNet. We replace ResNet with ResNeXt and implement a multi-branch structure for better low-level feature extraction. Additionally, we integrate SKNet attention mechanism with the C2f structure from YOLOv8, improving the model’s focus on critical image regions and enhancing the detection of minor defects. We also introduce an elliptical Gaussian kernel for size regression loss, better representing the aspect ratio of rail defects. This approach enhances detection accuracy and speeds up training. Our model achieves a mean accuracy (mAP) of 0.952 on the rail defects dataset, outperforming other models with a 6.6% improvement over the original and a 35.5% increase in training speed. These results demonstrate the efficiency and reliability of our method for rail defect detection.

Improving AI-assisted video editing: Optimized footage analysis through multi-task learning

In recent years, AI-assisted video editing has shown promising applications. Understanding and analyzing camera language accurately is fundamental in video editing, guiding subsequent editing and production processes. However, many existing methods for camera language analysis overlook computational efficiency and deployment requirements in favor of improving classification accuracy. Consequently, they often fail to meet the demands of scenarios with limited computing power, such as mobile devices. To address this challenge, this paper proposes an efficient multi-task camera language analysis pipeline based on shared representations. This approach employs a multi-task learning architecture with hard parameter sharing, enabling different camera language classification tasks to utilize the same low-level feature extraction network, thereby implicitly learning feature representations of the footage. Subsequently, each classification sub-task independently learns the high-level semantic information corresponding to the camera language type. This method significantly reduces computational complexity and memory usage while facilitating efficient deployment on devices with limited computing power. Furthermore, to enhance performance, we introduce a dynamic task priority strategy and a conditional dataset downsampling strategy. The experimental results demonstrate that achieved a comprehensive accuracy surpassing all previous methods. Moreover, training time was reduced by 66.33%, inference cost decreased by 59.85%, and memory usage decreased by 31.95% on the 2-task dataset MovieShots; on the 4-task dataset AVE, training time was reduced by 95.34%, inference cost decreased by 97.23%, and memory usage decreased by 61.21%.

AI-IMAGE REPRESENTATION AND LINEAR REPRENDER RENDERING

Image representation and rendering have become critical in numerous applications such as virtual reality, medical imaging, and computer graphics. Traditional rendering techniques often face challenges in efficiently handling complex scenes and achieving photorealistic results while maintaining low computational costs. The problem lies in the high-dimensional nature of image data, leading to slow processing times and reduced scalability. This research presents an AI-enhanced technique called Linear RepRender, which leverages deep learning to transform high-dimensional image representations into simplified linear forms for faster rendering. The proposed method employs a combination of convolutional neural networks (CNNs) and linear regression models to reduce image complexity. Specifically, the CNN extracts low-level and high-level features from the image, while the linear regression step approximates the scene’s core visual elements. This hybrid approach significantly improves rendering speed without sacrificing image quality. Furthermore, the method incorporates a loss function optimized for minimizing discrepancies between the rendered and ground truth images. Experimental results demonstrate that Linear RepRender outperforms traditional rendering algorithms, such as ray tracing and rasterization, in terms of computational efficiency and visual accuracy. On a dataset of complex 3D scenes, the proposed method achieved a 35% reduction in rendering time and a 22% improvement in peak signal-to-noise ratio (PSNR) compared to state-of-the-art methods. Additionally, Linear RepRender was able to handle up to 1.5 million polygons per scene with minimal visual artifacts, making it suitable for real-time applications.

Detecting Artificial Intelligence-Generated images via deep trace representations and interactive feature fusion

The detection of Artificial Intelligence-Generated (AIG) images plays an important role in verifying the authenticity and originality of digital images. However, recent advancements in state-of-the-art image generation methods have significantly challenged the ability to differentiate AIG images from natural photographs (NP). To address this issue, a novel approach based on deep trace representations and dual-branch interactive feature fusion is presented. Firstly, a global feature extraction module that leverages attention-based MobileViT (AT-MobileViT) is designed to learn the deep representations of the global trace information. Besides, we apply multiple enhanced residual blocks to extract discriminative multi-scale features. After that, a low-level feature extraction module incorporating a channel-spatial attention (CSA) block is also carefully employed to enhance the learning of trace representations. To facilitate the capture of complementary information between features, a dual-branch interactive feature fusion module is introduced by reshaping feature vectors into interactive matrices. By conducting experiments on both seen and unseen images, results demonstrate the better performance and robustness of the proposed method.

Rail Surface Defect Detection Based on Dual-Path Feature Fusion

With the rapid development of rail transit, the workload of track maintenance has increased, making the intelligent identification of rail surface defects crucial for improving detection efficiency. To address issues such as low defect detection accuracy, the loss of feature information due to single-path architecture backbones, and insufficient information interaction in existing rail defect detection methods, we propose a rail surface defect detection method based on dual-path feature fusion (DPF). This method initially employs a dual-path structure to separately extract low-level and high-level features. It then utilizes a combination of attention mechanisms and feature fusion techniques to integrate these features. By doing so, it preserves richer information and enhances detection accuracy and robustness. The experimental results demonstrate that the comprehensive performance of the proposed model is superior to mainstream algorithms.

Extraction of water bodies from high-resolution remote sensing imagery based on a deep semantic segmentation network

The precise delineation of urban aquatic features is of paramount importance in scrutinizing water resources, monitoring floods, and devising water management strategies. Addressing the challenge of indistinct boundaries and the erroneous classification of shadowed regions as water in high-resolution remote sensing imagery, we introduce WaterDeep, which is a novel deep learning framework inspired by the DeepLabV3 + architecture and an innovative fusion mechanism for high- and low-level features. This methodology first creates a comprehensive dataset of high-resolution remote sensing images, then progresses through the Xception baseline network for low-level feature extraction, and harnesses densely connected Atrous Spatial Pyramid Pooling (ASPP) modules to assimilate multi-scale data into sophisticated high-level features. Subsequently, the network decoder amalgamates the elemental and intricate features and applies dual-line interpolation to the amalgamated dataset to extract aqueous formations from the remote images. Experimental evidence substantiates that WaterDeep outperforms its existing deep learning counterparts, achieving a stellar overall accuracy of 99.284%, FWIoU of 95.58%, precision of 97.562%, recall of 95.486%, and F1 score of 96.513%. It also excels in the precise demarcation of edges and the discernment of shadows cast by urban infrastructure. The superior efficacy of the proposed method in differentiating water bodies in complex urban environments has significant practical applications in real-world contexts.

Improving Single-Image Super-Resolution with Dilated Attention

Single-image super-resolution (SISR) techniques have become a vital tool for improving image quality and clarity in the rapidly evolving field of digital imaging. Convolutional neural network (CNN) and transformer-based SISR techniques are very popular. However, CNN-based techniques are not suitable when capturing long-range dependencies, and transformer-based techniques suffer from computational complexity. To tackle these problems, this paper proposes a novel method called dilated attention-based single-image super-resolution (DAIR). It comprises three components: low-level feature extraction, multi-scale dilated transformer block (MDTB), and high-quality image reconstruction. A convolutional layer is used to extract the base features from low-resolution images, which lays the foundation for subsequent processing. Dilated attention is introduced to MDTB to enhance its ability to capture image features at different scales and ensure superior image details and structure recovery. After that, MDTB refines these features to extract multi-scale global attributes and effectively grasps images’ long-distance relationships and features across multiple scales. Finally, low-level features obtained from feature extraction and multi-scale global features obtained from MDTB are aggregated to reconstruct high-resolution images. The comparison with existing methods validates the efficacy of the proposed method and demonstrates its advantage in improving image resolution and quality.

HTC-retina: A hybrid retinal diseases classification model using transformer-Convolutional Neural Network from optical coherence tomography images

Retinal diseases are among nowadays major public health issues, deservedly needing advanced computer-aided diagnosis. We propose a hybrid model for multi label classification, whereby seven retinal diseases are automatically classified from Optical Coherence Tomography (OCT) images. We show that, by combining the strengths of Convolutional Neural Networks (CNNs) and Visual Transformers (ViTs), we can produce a more powerful type of model for medical image classification, especially when considering local lesion information such as retinal diseases. CNNs are indeed proved to be efficient at parameter utilization and provide the ability to extract local features and multi-scale feature maps through convolutional operations. On the other hand, ViT’s self-attention procedure allows processing long-range and global dependencies within an image. The paper clearly shows that the hybridization of these complementary capabilities (CNNs-ViTs) presents a high image processing potential that is more robust and efficient. The proposed model adopts a hierarchical CNN module called Convolutional Patch and Token Embedding (CPTE) instead of employing a direct tokenization approach using the raw input OCT image in the transformer. The CPTE module’s role is to incorporate an inductive bias, to reduce the reliance on large-scale datasets, and to address the low-level feature extraction challenges of the ViT. In addition, considering the importance of local lesion information in OCT images, the model relies on a parallel module called Residual Depthwise-Pointwise ConvNet (RDP-ConvNet) for extracting high-level features. RDP-ConvNet utilizes depthwise and pointwise convolution layers within a residual network architecture. The overall performance of the HTC-Retina model was evaluated on three datasets: the OCT-2017, OCT-C8, and OCT-2014 ; outperforming previous established models, achieving accuracy rates of 99.40%, 97.00%, and 99.77%, respectively ; and sensitivity rates of 99.41%, 97.00%, and 99.77%, respectively. Notably, the model showed high performance while maintaining computational efficiency.

Dual-stream mutually adaptive quality assessment for authentic distortion image

To address semantic misperception caused by distortion and accurately simulate human perceptual processes, this study proposed a method that utilised dual-stream mutually adaptive feature mapping. To extract complementary features, a dual-stream unsupervised method was used during the training phase. One stream was responsible for the extraction of low-level and global features, whereas the other was dedicated to extracting high-level semantic and positional features. Following the freezing of feature extraction network weights, we proposed a structure that used standard deviation labels to predict the quality distribution. The experimental results obtained from 10 published image quality databases demonstrated the superiority of the proposed algorithm. The algorithm outperformed the majority of mainstream methods when evaluated on authentic distortion databases and exhibited competitive performance on synthetic distortion databases.

Finetuning foundation models for joint analysis optimization in High Energy Physics

In this work we demonstrate that significant gains in performance and data efficiency can be achieved in High Energy Physics (HEP) by moving beyond the standard paradigm of sequential optimization or reconstruction and analysis components. We conceptually connect HEP reconstruction and analysis to modern machine learning workflows such as pretraining, finetuning, domain adaptation and high-dimensional embedding spaces and quantify the gains in the example usecase of searches of heavy resonances decaying via an intermediate di-Higgs system to four b-jets. To our knowledge this is the first example of a low-level feature extraction network finetuned for a downstream HEP analysis objective.

Automated Prediction of Malignant Melanoma using Two-Stage Convolutional Neural Network.

A skin lesion refers to an area of the skin that exhibits anomalous growth or distinctive visual characteristics compared to the surrounding skin. Benign skin lesions are noncancerous and generally pose no threat. These irregular skin growths can vary in appearance. On the other hand, malignant skin lesions correspond to skin cancer, which happens to be the most prevalent form of cancer in the United States. Skin cancer involves the unusual proliferation of skin cells anywhere on the body. The conventional method for detecting skin cancer is relatively more painful. This work involves the automated prediction of skin cancer and its types using two stage Convolutional Neural Network (CNN). The first stage of CNN extracts low level features and second stage extracts high level features. Feature selection is done using these two CNN and ABCD (Asymmetry, Border irregularity, Colour variation, and Diameter) technique. The features extracted from the two CNNs are fused with ABCD features and fed into classifiers for the final prediction. The classifiers employed in this work include ensemble learning methods such as gradient boosting and XG boost, as well as machine learning classifiers like decision trees and logistic regression. This methodology is evaluated using the International Skin Imaging Collaboration (ISIC) 2018 and 2019 dataset. As a result, the first stage CNN which is used for creation of new dataset achieved an accuracy of 97.92%. Second stage CNN which is used for feature selection achieved an accuracy of 98.86%. Classification results are obtained for both with and without fusion of features. Therefore, two stage prediction model achieved better results with feature fusion.

GLIMS: Attention-guided lightweight multi-scale hybrid network for volumetric semantic segmentation

Convolutional Neural Networks (CNNs) have become widely adopted for medical image segmentation tasks, demonstrating promising performance. However, the inherent inductive biases in convolutional architectures limit their ability to model long-range dependencies and spatial correlations. While recent transformer-based architectures address these limitations by leveraging self-attention mechanisms to encode long-range dependencies and learn expressive representations, they often struggle to extract low-level features and are highly dependent on data availability. This motivated us for the development of GLIMS, a data-efficient attention-guided hybrid volumetric segmentation network. GLIMS utilizes Dilated Feature Aggregator Convolutional Blocks (DACB) to capture local–global feature correlations efficiently. Furthermore, the incorporated Swin Transformer-based bottleneck bridges the local and global features to improve the robustness of the model. Additionally, GLIMS employs an attention-guided segmentation approach through Channel and Spatial-Wise Attention Blocks (CSAB) to localize expressive features for fine-grained border segmentation. Quantitative and qualitative results on glioblastoma and multi-organ CT segmentation tasks demonstrate GLIMS’ effectiveness in terms of complexity and accuracy. GLIMS demonstrated outstanding performance on BraTS2021 and BTCV datasets, surpassing the performance of Swin UNETR. Notably, GLIMS achieved this high performance with a significantly reduced number of trainable parameters. Specifically, GLIMS has 47.16 M trainable parameters and 72.30G FLOPs, while Swin UNETR has 61.98 M trainable parameters and 394.84G FLOPs. The code is publicly available at https://github.com/yaziciz/GLIMS.