High-level Features Research Articles

Zero shot plant disease classification with semantic attributes

In the rapidly evolving field of plant disease detection, the number and complexity of crop diseases are increasing, made worse by factors like climate change. Addressing these challenges requires robust and efficient methodologies capable of early and accurate disease identification. This paper explores the integration of advanced deep learning techniques, including pre-trained models, zero-shot learning, and semantic attributes to enhance the effectiveness of plant disease detection systems. High level features extracted from the images by these pretrained models capture crucial patterns, while domain-specific semantic attributes, such as leaf texture and color variations, enhance the understanding. Incorporating zero-shot learning enables adaptation to new and unseen diseases using semantic descriptions. Experimental validation across diverse plant species and disease types underscores the approach’s reliability in real-world agricultural scenarios. Our approach has demonstrated superior performance with plant village dataset, showing a significant improvement in accuracy and generalization. These results underscore the potential of our method to revolutionize plant disease detection and management in agricultural practices.

Attention-Guided Sample-Based Feature Enhancement Network for Crowded Pedestrian Detection Using Vision Sensors.

Occlusion presents a major obstacle in the development of pedestrian detection technologies utilizing computer vision. This challenge includes both inter-class occlusion caused by environmental objects obscuring pedestrians, and intra-class occlusion resulting from interactions between pedestrians. In complex and variable urban settings, these compounded occlusion patterns critically limit the efficacy of both one-stage and two-stage pedestrian detectors, leading to suboptimal detection performance. To address this, we introduce a novel architecture termed the Attention-Guided Feature Enhancement Network (AGFEN), designed within the deep convolutional neural network framework. AGFEN improves the semantic information of high-level features by mapping it onto low-level feature details through sampling, creating an effect comparable to mask modulation. This technique enhances both channel-level and spatial-level features concurrently without incurring additional annotation costs. Furthermore, we transition from a traditional one-to-one correspondence between proposals and predictions to a one-to-multiple paradigm, facilitating non-maximum suppression using the prediction set as the fundamental unit. Additionally, we integrate these methodologies by aggregating local features between regions of interest (RoI) through the reuse of classification weights, effectively mitigating false positives. Our experimental evaluations on three widely used datasets demonstrate that AGFEN achieves a 2.38% improvement over the baseline detector on the CrowdHuman dataset, underscoring its effectiveness and potential for advancing pedestrian detection technologies.

Color and Attention for U : Modified Multi Attention U-Net for a Better Image Colorization

Image colorization is a tedious task that requires creativity and understanding of the image context and semantic information. Many models have been made by harnessing various deep learning architectures to learn the plausible colorization. With the rapid discovery of new architecture and image generation techniques, more powerful options can be explored and improved for image colorization tasks. This research explores a new architecture to colorize an image by using pre-trained embeddings on U-Net combined with several attention modules across the model. Using embeddings from a pre-trained classifier provides a high-level feature extraction from the image. Conversely, multi-attention gives a little taste of image segmentation so that the model can distinguish objects in the image and further enhance the additional information given by the pre-trained embeddings. Adversarial training is also utilized as a normalization to make the generated image more realistic. This research preferred Parch GAN over base GAN as the discriminator model to ensure that the colorization has a consistent quality across all patches. The study shows that this U-Net modification can improve the generated image quality compared to a normal U-Net. The proposed architecture reaches an FID of 48.6253, SSIM of 0.8568, and PSNR of 19.7831 by only training it for 25 epochs; hence, this research offers another view of image colorization by using modules that are often used for image segmentation tasks.

BCL-Former: Localized Transformer Fusion with Balanced Constraint for polyp image segmentation

Polyp segmentation remains challenging for two reasons: (a) the size and shape of colon polyps are variable and diverse; (b) the distinction between polyps and mucosa is not obvious. To solve the above two challenging problems and enhance the generalization ability of segmentation method, we propose the Localized Transformer Fusion with Balanced Constraint (BCL-Former) for Polyp Segmentation. In BCL-Former, the Strip Local Enhancement module (SLE module) is proposed to capture the enhanced local features. The Progressive Feature Fusion module (PFF module) is presented to make the feature aggregation smoother and eliminate the difference between high-level and low-level features. Moreover, the Tversky-based Appropriate Constrained Loss (TacLoss) is proposed to achieve the balance and constraint between True Positives and False Negatives, improving the ability to generalize across datasets. Extensive experiments are conducted on four benchmark datasets. Results show that our proposed method achieves state-of-the-art performance in both segmentation precision and generalization ability. Also, the proposed method is 5%–8% faster than the benchmark method in training and inference. The code is available at: https://github.com/sjc-lbj/BCL-Former.

A novel BH3DNet method for identifying pine wilt disease in Masson pine fusing UAS hyperspectral imagery and LiDAR data

Pine Wilt Disease (PWD) is a forest infectious disease that inflicts substantial economic losses to China’s forestry. Its rapid spread and the significant challenges associated with its control make early detection of infected trees crucial for disaster prevention. Unmanned aerial systems (UASs) hyperspectral imaging (HSI) and light detection and ranging (LiDAR) technologies provide high-resolution spectral diagnostic information coupled with intricate three-dimensional structural data, which has potential for fine grained monitoring of PWD. However, how to fuse HSI and LiDAR data to identify the early infected individual trees is still a challenge. This study presents a novel instance segmentation network, BH3DNet, to identify individual trees at different PWD-infected stages by extracting high-level abstract features based on the fusion of drone HSI and LiDAR data. BH3DNet introduces the PointNet++ model as the base network, and incorporates a shared encoder and twin parallel decoders to align semantic category prediction and instance segmentation of individual trees in an end-to-end approach. By applying an enhanced point cloud dataset that fuses drone HSI and LiDAR point cloud data, this model facilitates the identification of PWD infection stages at the individual tree scale. We evaluated the proposed model in a Masson pine forest stand sparsely mixed with broadleaf trees in a variety of infection states ranging from healthy to severely infected by PWD, and compared the performance of the model using the RGB bands, full HSI bands and screened bands as inputs, respectively. BH3DNet achieves an overall accuracy of 89.65 % with a Kappa × 100 of 87.29 for identifying individual trees using screened HSI bands and LiDAR point cloud, significantly outperforming the Mask R-CNN using only HSI data (overall accuracy: 70.81 %, Kappa × 100: 64.16). Moreover, BH3DNet’s accuracy at the early infection stage reaches 83.75 %. It proves that fusing HSI and point cloud data reflects the information of individual trees distribution and infection status, and the BH3DNet is suitable for high-precision monitoring of PWD.

Optimizing feature extraction and fusion for high-resolution defect detection in solar cells

In this paper, we propose a novel architecture for defect detection in electroluminescent images of polycrystalline silicon solar cells, addressing the challenges posed by subtle and dispersed defects. Our model, based on a modified Swin Transformer, incorporates key innovations that enhance feature extraction and fusion. We replace the conventional self-attention mechanism with a novel group self-attention mechanism, increasing the mAP50:5:95 score from 50.12 % to 52.98 % while reducing inference time from 74 ms to 62 ms. We also introduce a spatial displacement with shift convolution module, replacing the traditional Multi-Layer Perceptron, which further enhances the model's receptive field and improves precision and recall. Additionally, our fast multi-scale feature fusion mechanism effectively combines high-resolution details with high-level semantic features from different network layers, optimizing defect detection accuracy. Experimental results on the PVEL-AD dataset demonstrate that our model achieves the highest mAP50 score of 83.11 % and an F1-Score of 84.33 %, surpassing state-of-the-art models while maintaining a competitive inference time of 66.3 ms. These findings highlight the effectiveness of our innovations in improving defect detection accuracy and computational efficiency, making our model a robust solution for quality assurance in solar cell manufacturing.

MGCN-PolyA: An Integrated Computational Framework for Predicting Poly(A) Signals with Multiscale-gated Convolutional Networks

Background: The accurate recognition of the polyadenylation signal (PAS) from DNA sequences is essential for understanding gene transcriptional regulation. A variety of machine learning-based computational methods have been developed to predict PAS in recent years; however, their performance and their generalization ability are unsatisfactory. It is highly desirable to design more preferable computational approaches for PAS prediction. Methods: In this work, we developed an integrated framework MGCN-PolyA for PAS prediction across four species, including Homo sapiens, Bos taurus, Mus musculus, and Drosophila melanogaster. MGCN-Poly(A) benefits from the diversity of feature engineering and the effectiveness of the model architecture. We combined features from different perspectives, such as word embedding, One-hot encoding, K-mer frequency, and Enhanced Nucleic Acid Composition (ENAC), which complement each other and provide rich and comprehensive information for model learning. In model architecture, MGCN-Poly(A) leverages a two-channel multi-scale gated convolutional network to effectively learn high-level feature representations at different scales, and then combines the statistical features to predict PAS using random forest algorithm. These designs not only speed up network training, but also improves the generalization ability Results: The benchmarking experiments on the independent test datasets demonstrate that MGCNPolyA outperforms other state-of-the-art algorithms in identifying PAS. MGCN-PolyA has the highest accuracy on all test datasets, and its excellent performance on cross-species validation also demonstrates the robustness of our model. Conclusion: Extracting features from different perspectives is important for PAS recognition, and the integration of DNNs and shallow machine learning algorithms can improve the model performance.

Genetic Programming for Automatically Evolving Multiple Features to Classification

Abstract Performing classification on high-dimensional data poses a significant challenge due to the huge search space. Moreover, complex feature interactions introduce an additional obstacle. The problems can be addressed by using feature selection to select relevant features or feature construction to construct a small set of high-level features. However, performing feature selection or feature construction only might make the feature set suboptimal. To remedy this problem, this study investigates the use of genetic programming for simultaneous feature selection and feature construction in addressing different classification tasks. The proposed approach is tested on 16 datasets and compared with seven methods including both feature selection and feature constructions techniques. The results show that the obtained feature sets with the constructed and/or selected features can significantly increase the classification accuracy and reduce the dimensionality of the datasets. Further analysis reveals the complementarity of the obtained features leading to the promising classification performance of the proposed method.

A Convolutional Neural Network Approach for Precision Fish Disease Detection

Background: Detecting and classifying fish diseases is crucial for maintaining the health and sustainability of aquaculture systems. This study employs deep learning techniques, particularly Convolutional Neural Networks (CNNs), to automate the detection of various fish diseases using image data. Methods: The study utilizes a carefully curated dataset sourced from the Kaggle database, comprising images representing seven distinct types of fish diseases, along with images of healthy fish. Data preprocessing techniques, including resizing, rescaling, denoising, sharpening, and smoothing, are applied to enhance image quality and facilitate accurate disease detection. Data augmentation is employed to increase the model's ability to generalize to unseen data. The CNN architecture is designed with cascading convolutional layers, ReLU activation functions, and pooling operations to extract high-level features associated with fish infections. The model architecture, implemented using the Keras Sequential API, includes convolutional layers, max pooling layers, and densely connected layers for classification. Results: Experimental results demonstrated promising performance across various disease categories, with high accuracy and balanced precision and recall values for most classes. The study also discussed the impact of climate change on fish disease incidence and underscores the importance of effective monitoring and management practices facilitated by technological innovations such as Big Data, IoT, sensors, and robotics in ensuring sustainable fisheries management.

TPAFNet: Transformer-Driven Pyramid Attention Fusion Network for 3D Medical Image Segmentation.

The field of 3D medical image segmentation is witnessing a growing trend in the utilization of combined networks that integrate convolutional neural networks and transformers. Nevertheless, prevailing hybrid networks are confronted with limitations in their straightforward serial or parallel combination methods and lack an effective mechanism to fuse channel and spatial feature attention. To address these limitations, we present a robust multi-scale 3D medical image segmentation network, the Transformer-Driven Pyramid Attention Fusion Network, which is denoted as TPAFNet, leveraging a hybrid structure of CNN and transformer. Within this framework, we exploit the characteristics of atrous convolution to extract multi-scale information effectively, thereby enhancing the encoding results of the transformer. Furthermore, we introduce the TPAF block in the encoder to seamlessly fuse channel and spatial feature attention from multi-scale feature inputs. In contrast to conventional skip connections that simply concatenate or add features, our decoder is enriched with a TPAF connection, elevating the integration of feature attention between low-level and high-level features. Additionally, we propose a low-level encoding shortcut from the original input to the decoder output, preserving more original image features and contributing to enhanced results. Finally, the deep supervision is implemented using a novel CNN-based voxel-wise classifier to facilitate better network convergence. Experimental results demonstrate that TPAFNet significantly outperforms other state-of-the-art networks on two public datasets, indicating that our research can effectively improve the accuracy of medical image segmentation, thereby assisting doctors in making more precise diagnoses.

Dental Caries Segmentation using Deformable Dense Residual Half U-Net for Teledentistry System

Clinical practitioners’ workload and challenges are significantly reduced by classifying, predicting, and localizing lesions or dental caries. In recent research, a high-reliability diagnostic system within deep learning models has been implemented in a clinical teledentistry system. In order to construct an efficient, precise, and lightweight deep learning architecture, it is dynamically structured. In this paper, we present an efficient, accurate, and lightweight deep learning architecture for augmenting spatial locations and improving the transformation modeling abilities of fixed-structure CNNs. Deformable Dense Residual (DDR) enhances the efficacy of the residual convolution block by optimizing its structure, thereby mitigating model redundancy and ameliorating the challenge of vanishing gradients encountered during the training stages. DDR Half U-Net presents notable advancements to the simplified U-Net framework across three pivotal domains: the encoder, decoder, and loss function. Specifically, the encoder integrates deformable convolutions, thereby enhancing the model's capacity to discern features of diverse scales and configurations. In the decoder, a sophisticated arrangement of dense residual connections facilitates the fusion of low-level and high-level features, contributing to comprehensive feature extraction. Moreover, the utilization of a weight-adaptive loss function ensures equitable consideration of both caries and non-caries samples, thereby promoting balanced optimization during training.

Lung Cancer Detection Using Convolutional Neural Networks

Lung cancer is among the most prevalent and deadly cancers worldwide. Accurate diagnosis and early detection are critical for improving lung cancer patient outcomes and survival rates. Thanks to developments in medical imaging technology, computer-aided diagnosis (CAD) systems have shown a great deal of promise in helping radiologists identify and diagnose lung cancer from medical images. Here, we present the use of convolutional neural networks (CNNs) to create an early detection system (CAD) for lung cancer. The suggested approach uses lung computed tomography (CT) scans as input and use a CNN architecture to extract high-level features from the pictures. We use transfer learning to enhance a CNN model trained on a large dataset of CT images. The CNN model has been taught to determine if a specific CT image contains lung cancer or not. We evaluate the performance of the proposed CAD system on a dataset of CT scans of the lungs from different institutions. The trial's results show that our CNN-based CAD system can reliably and precisely identify lung cancer from CT scans. We also show the comparative performance of our proposed system against the state-of-the-art machine learning methods for lung cancer prediction. In conclusion, the suggested CNN-based deep learning-based CAD system has produced encouraging results for lung cancer detection from CT scans. The approach might help radiologists identify and classify lung cancer early on, leading to better patient outcomes and survival rates. The viability and usefulness of the suggested approach in clinical practice require more study

Continuous blood pressure monitoring based on transformer encoders and stacked attention gated recurrent units

Continuous blood pressure monitoring (CBPM) is critical to support the accurate prevention and reliable treatment of cardiovascular diseases. To achieve efficient multi-information interaction and further improve the monitoring performance, this research proposes an intelligent model based on transformer encoders and stacked attention gated recurrent units (TE-SAGRU) for CBPM. Long-term multi-source feature sequences with rich information are initially extracted from photoplethysmography (PPG) and electrocardiography (ECG) signals. The paralleled transformer encoders are constructed for different source feature sequences to obtain high-level feature representations and preserve respective long-term independence. The multiple stacked attention gated recurrent units are cross-connected for multi-interactive feature fusion and promoting complementarity effects of multi-source features on CBPM. Comprehensive comparison experiments are carried out to validate the effectiveness of the TE-SAGRU model, using the dataset with 1000 subjects derived from the MIMIC-III database. The continuous monitoring errors of the TE-SAGRU model for systolic blood pressure (SBP) and diastolic blood pressure (DBP) are 3.91 ± 5.65 mmHg and 2.29 ± 3.01 mmHg. The monitored results pass the requirement of the Association for the Advancement of Medical Instrumentation (AAMI) standard and achieve Grade A of the British Hypertension Society (BHS) protocol.

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).

Detecting floating litter in freshwater bodies with semi-supervised deep learning

Researchers and practitioners have extensively utilized supervised Deep Learning methods to quantify floating litter in rivers and canals. These methods require the availability of large amount of labeled data for training. The labeling work is expensive and laborious, resulting in small open datasets available in the field compared to the comprehensive datasets for computer vision, e.g., ImageNet. Fine-tuning models pre-trained on these larger datasets helps improve litter detection performances and reduces data requirements. Yet, the effectiveness of using features learned from generic datasets is limited in large-scale monitoring, where automated detection must adapt across different locations, environmental conditions, and sensor settings. To address this issue, we propose a two-stage semi-supervised learning method to detect floating litter based on the Swapping Assignments between multiple Views of the same image (SwAV). SwAV is a self-supervised learning approach that learns the underlying feature representation from unlabeled data. In the first stage, we used SwAV to pre-train a ResNet50 backbone architecture on about 100k unlabeled images. In the second stage, we added new layers to the pre-trained ResNet50 to create a Faster R-CNN architecture, and fine-tuned it with a limited number of labeled images (≈1.8k images with 2.6k annotated litter items). We developed and validated our semi-supervised floating litter detection methodology for images collected in canals and waterways of Delft (the Netherlands) and Jakarta (Indonesia). We tested for out-of-domain generalization performances in a zero-shot fashion using additional data from Ho Chi Minh City (Vietnam), Amsterdam and Groningen (the Netherlands). We benchmarked our results against the same Faster R-CNN architecture trained via supervised learning alone by fine-tuning ImageNet pre-trained weights. The findings indicate that the semi-supervised learning method matches or surpasses the supervised learning benchmark when tested on new images from the same training locations. We measured better performances when little data (≈200 images with about 300 annotated litter items) is available for fine-tuning and with respect to reducing false positive predictions. More importantly, the proposed approach demonstrates clear superiority for generalization on the unseen locations, with improvements in average precision of up to 12.7%. We attribute this superior performance to the more effective high-level feature extraction from SwAV pre-training from relevant unlabeled images. Our findings highlight a promising direction to leverage semi-supervised learning for developing foundational models, which have revolutionized artificial intelligence applications in most fields. By scaling our proposed approach with more data and compute, we can make significant strides in monitoring to address the global challenge of litter pollution in water bodies.

Tissue segmentation for traumatic brain injury based on multimodal MRI image fusion-semantic segmentation

Accurate segmentation of traumatic brain injury (TBI) has great significance for physicians to diagnose and assess a patient’s condition. The utilization of multimodal information plays a critical role in TBI segmentation. However, most of the existing methods mainly focus on direct extraction and selection of deep semantic features, whereas in this paper, we use image fusion as an auxiliary task for feature learning based on multimodal feature extraction to achieve more sufficient fusion of multimodal features. Therefore, we design a multimodal image fusion-semantic segmentation based framework. The proposed approach mainly consists of a semantic encoder module, a semantic segmentation module and an image fusion module. The semantic encoder compresses the input image into a smaller feature space to extract semantic features. The semantic segmentation module utilizes both the detailed information extracted by the encoder and the semantic information of high-level features extracted from the semantic segmentation module to generate the segmentation results. The image fusion module fuses semantic feature information from different modalities as an auxiliary task to semantic segmentation. Furthermore, to enhance the model’s performance even further, an uncertainty-based approach is employed, which dynamically adjusts the loss weights for the image fusion task and the semantic segmentation task during the model training process. The proposed method is evaluated on a private dataset, and compared with other widely recognized methods. It demonstrates outstanding performance in both Dice score and Recall metrics.

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.

A Cross-Level Iterative Subtraction Network for Camouflaged Object Detection

Camouflaged object detection (COD) is a challenging task, aimed at segmenting objects that are similar in color and texture to their background. Sufficient multi-scale feature fusion is crucial for accurately segmenting object regions. However, most methods usually focus on information compensation, overlooking the difference between features, which is important for distinguishing the object from the background. To this end, we propose the cross-level iterative subtraction network (CISNet), which integrates information from cross-layer features and enhances details through iteration mechanisms. CISNet involves a cross-level iterative structure (CIS) for feature complementarity, where texture information is used to enrich high-level features and semantic information is used to enhance low-level features. In particular, we present a multi-scale strip convolution subtraction (MSCSub) module within CIS to extract difference information between cross-level features and fuse multi-scale features, which improves the feature representation and guides accurate segmentation. Furthermore, an enhanced guided attention (EGA) module is presented to refine features by deeply mining local context information and capturing a broader range of relationships between different feature maps in a top-down manner. Extensive experiments conducted on four benchmark datasets demonstrate that our model outperforms the state-of-the-art COD models in all evaluation metrics.

An Improved YOLOv5 Model for Concrete Bubble Detection Based on Area K-Means and ECANet

The appearance quality of fair-faced concrete plays a crucial role in evaluating the engineering quality, as the abundance of small-area bubbles generated during construction diminishes the surface quality of concrete. However, existing methods are plagued by sluggish detection speed and inadequate accuracy. Therefore, this paper proposes an improved method based on YOLOv5 to rapidly and accurately detect small bubble defects on the surface of fair-faced concrete. Firstly, to address the issue of YOLOv5 in generating prior boxes for imbalanced samples, we divide the image preprocessing part into small-, medium-, and large-area intervals corresponding to the number of heads. Additionally, we propose an area-based k-means clustering approach specifically tailored for the anchor boxes within each of these intervals. Moreover, we adjust the number of prior boxes generated by k-means clustering according to the training loss function to adapt to bubbles of different sizes. Then, we introduce the ECA (Efficient Channel Attention) mechanism into the neck part of the model to effectively capture inter-channel interactions and enhance feature representation. Subsequently, we incorporate feature concatenation in the neck part to facilitate the fusion of low-level and high-level features, thereby improving the accuracy and generalization ability of the network. Finally, we construct our own dataset containing 980 images of two classes: cement and bubbles. Comparative experiments are conducted on our dataset using YOLOv5s, YOLOv6s, YOLOxs, and our method. Experimental results demonstrate that the proposed method achieves the highest detection accuracy in terms of mAP0.5, mAP0.75, and mAP0.5:0.95. Compared to YOLOv5s, our method achieves a 7.1% improvement in mAP0.5, a 3.7% improvement in mAP0.75, and a 4.5% improvement in mAP0.5:0.95.

RFSC-net: Re-parameterization forward semantic compensation network in low-light environments

Although detectors currently perform well in well-light conditions, their accuracy decreases due to insufficient object information. In addressing this issue, we propose the Re-parameterization Forward Semantic Compensation Network (RFSC-Net). We propose the Reparameterization Residual Efficient Layer Aggregation Networks (RSELAN) for feature extraction, which integrates the concepts of re-parameterization and the Efficient Layer Aggregation Networks (ELAN). While focusing on the fusion of feature maps of the same dimension, it also incorporates upward fusion of lower-level feature maps, enhancing the detailed texture information in higher-level features. Our proposed Forward Semantic Compensation Feature Fusion (FSCFF) network reduces interference from high-level to low-level semantic information, retaining finer details to improve detection accuracy in low-light conditions. Experiments on the low-light ExDark and DarkFace datasets show that RFSC-Net improves mAP by 2% on ExDark and 0.5% on DarkFace over the YOLOv8n baseline, without an increase in parameter counts. Additionally, AP50 is enhanced by 2.1% on ExDark and 1.1% on DarkFace, with a mere 3.7 ms detection latency on ExDark.