Object Detection Framework Research Articles

SOD-YOLO: A lightweight small object detection framework

Currently, lightweight small object detection algorithms for unmanned aerial vehicles (UAVs) often employ group convolutions, resulting in high Memory Access Cost (MAC) and rendering them unsuitable for edge devices that rely on parallel computing. To address this issue, we propose the SOD-YOLO model based on YOLOv7, which incorporates a DSDM-LFIM backbone network and includes a small object detection branch. The DSDM-LFIM backbone network, which combines Deep-Shallow Downsampling Modules (DSD Modules) and Lightweight Feature Integration Modules (LFI Modules), avoids excessive use of group convolutions and element-wise operations. The DSD Module focuses on extracting both deep and shallow features from feature maps using fewer parameters to obtain richer feature representations. The LFI Module, is a dual-branch feature integration module designed to consolidate feature information. Experimental results demonstrate that the SOD-YOLO model achieves an AP50 of 50.7% and a FPS of 72.5 on the VisDrone validation set. Compared to YOLOv7, our model reduces computational costs by 20.25% and decreases the number of parameters by 17.89%. After scaling the number of channels in the model, it achieves an AP50 of 33.4% with an inference time of 27.3ms on the Atlas 200I DK A2. These experimental results indicate that the SOD-YOLO model can effectively perform small object detection tasks in a large number of aerial images captured by UAVs.

Road Defect Identification and Location Method Based on an Improved ML-YOLO Algorithm

The conventional method for detecting road defects relies heavily on manual inspections, which are often inefficient and struggle with precise defect localization. This paper introduces a novel approach for identifying and locating road defects based on an enhanced ML-YOLO algorithm. By refining the YOLOv8 object detection framework, we optimize both the traditional convolutional layers and the spatial pyramid pooling network. Additionally, we incorporate the Convolutional Block Attention to effectively capture channel and spatial features, along with the Selective Kernel Networks that dynamically adapt to feature extraction across varying scales. An optimized target localization algorithm is proposed to achieve high-precision identification and accurate positioning of road defects. Experimental results indicate that the detection accuracy of the improved ML-YOLO algorithm reaches 0.841, with a recall rate of 0.745 and an average precision of 0.817. Compared to the baseline YOLOv8 model, there is an increase in accuracy by 0.13, a rise in recall rate by 0.117, and an enhancement in average precision by 0.116. After the high detection accuracy of road defects was confirmed, generalization experiments were carried out on the improved ML-YOLO model in the public data set. The experimental results showed that compared with the original YOLOv8n, the average precision and recall rate of all types of ML-YOLO increased by 0.075, 0.121, and 0.035 respectively, indicating robust generalization capabilities. When applied to real-time road monitoring scenarios, this algorithm facilitates precise detection and localization of defects while significantly mitigating traffic accident risks and extending roadway service life. A high detection accuracy of road defects was achieved.

Salient object detection with bayesian inference based on radar and camera fusion used in UAV obstacle avoidance

Most existing salient object detection methods are sensitive to background noise and rely on prior information in UAV obstacle avoidance applications despite detection methods witnessing rapid progress. In this paper, we propose an efficient framework for salient object detection based on radar-camera fusion and iterative Bayesian optimization. A rough salient object (RSO) image is first built through radar and camera calibration. Next, the RSO image is used to calculate the edge response based on the receptive field mechanism of the primary visual cortex to construct the contour image. Finally, the above two images and the 2D Gaussian distribution are jointly integrated into an iterative Bayesian optimization scheme to get the final salient object image. Different from typical detection methods, this method suppresses background noise by filtering out irrelevant pixels using fusion information. The Bayesian inference framework’s detection performance is improved by precise spatial prior, consisting of optimized contour and RSO images. Experimental results indicate that the presented algorithm performs well against state-of-the-art salient object detection methods on several reference datasets in different evaluation metrics.

Video Content Analysis for Detection of Road Traffic Congestion Pattern: An Optimized YOLOv8 Based Approach

This paper presents a new approach to Video Content Analysis (VCA) with a specific focus on detecting and analyzing road traffic congestion patterns. The implemented methodology deploys the YOLOv8 (You Only Look Once) object detection framework, which has been optimized to accurately and efficiently identify road traffic objects within video streams. The primary goal is to improve traditional VCA systems by capturing and understanding complex road congestion scenarios. This research develops a customized YOLO model optimized specifically for road traffic analysis. The optimization process addresses the challenges inherent in real-world traffic scenarios including variations in lighting conditions, diverse vehicle types and dynamic traffic flow patterns. The developed model seeks to enable robust and real-time detection of congestion-related events such as heavy congestion and normal congestion thereby contributing to improved traffic management and overall road safety. Moreover, the paper explores the integration of advanced techniques for congestion pattern analysis including vehicle detection, tracking, and counting. The combination of these features facilitates a comprehensive understanding of traffic dynamics and congestion evolution over time. The proposed approach is trained and evaluated on benchmark datasets and real-world video footage to assess its performance. Experimental results demonstrate the effectiveness of the optimized YOLOv8-based approach in accurately detecting and analyzing road traffic congestion patterns. The system exhibits promising capabilities in handling diverse and challenging scenarios thereby serving as a valuable tool for intelligent transportation systems, urban planning and traffic management. This research significantly contributes to the advancement of video-based traffic analysis, providing a robust solution for monitoring and mitigating congestion-related challenges on road networks.

An Improved YOLOv8-Based Foreign Detection Algorithm for Transmission Lines

This research aims to overcome three major challenges in foreign object detection on power transmission lines: data scarcity, background noise, and high computational costs. In the improved YOLOv8 algorithm, the newly introduced lightweight GSCDown (Ghost Shuffle Channel Downsampling) module effectively captures subtle image features by combining 1 × 1 convolution and GSConv technology, thereby enhancing detection accuracy. CSPBlock (Cross-Stage Partial Block) fusion enhances the model’s accuracy and stability by strengthening feature expression and spatial perception while maintaining the algorithm’s lightweight nature and effectively mitigating the issue of vanishing gradients, making it suitable for efficient foreign object detection in complex power line environments. Additionally, PAM (pooling attention mechanism) effectively distinguishes between background and target without adding extra parameters, maintaining high accuracy even in the presence of background noise. Furthermore, AIGC (AI-generated content) technology is leveraged to produce high-quality images for training data augmentation, and lossless feature distillation ensures higher detection accuracy and reduces false positives. In conclusion, the improved architecture reduces the parameter count by 18% while improving the mAP@0.5 metric by a margin of 5.5 points when compared to YOLOv8n. Compared to state-of-the-art real-time object detection frameworks, our research demonstrates significant advantages in both model accuracy and parameter size.

A NOVEL HYBRID DEEP LEARNING APPROACH FOR 3D OBJECT DETECTION AND TRACKING IN AUTONOMOUS DRIVING

Recently Object detection and tracking using fusion of LiDAR and RGB camera for the autonomous vehicle environment is a challenging task. The existingworks initiates several object detection and tracking frameworks using ArtificialIntelligence (AI) algorithms. However, they were limited with high false positives and computation time issues thus lacking the performance of autonomousdriving environment. The existing issues are resolved by proposing HybridDeep Learning based Multi Object Detection and Tracking (HDL-MODT) using sensor fusion methods. The proposed work performs fusion of solid stateLiDAR, Pseudo LiDAR, and RGB camera for improving detection and trackingquality. At first, the multi-stage preprocessing is done in which noise removal isperformed using Adaptive Fuzzy Filter (A-Fuzzy). The pre-processed fused image is then provided for instance segmentation to reduce the classification andtracking complexity. For that, the proposed work adopts Lightweight GeneralAdversarial Networks (LGAN). The segmented image is provided for objectdetection and tracking using HDL. For reducing the complexity, the proposedwork utilized VGG-16 for feature extraction which forms the feature vectors.The features vectors are then provided for object detection using YOLOv4.Finally, the detected objects were tracked using Improved Unscented KalmanFilter (IUKF) and mapping the vehicles using time based mapping by considering their RFID, velocity, location, dimension and unique ID. The simulation ofthe proposed work is carried out using MATLAB R2020a simulation tool andperformance of the proposed work is compared with several metrics that showthat the proposed work outperforms than the existing works.

Beyond clean data: Exploring the effects of label noise on object detection performance

In recent years, the growth of large-scale datasets has significantly propelled the progress of deep learning applications. Yet, annotating these datasets remains a labor-intensive endeavor, pushing the reliance on cost-effective but less specialized data collection methods and internet data sources. This often results in noisy and inaccurate labels, compromising data quality. Traditional machine learning models assume clean data, but real-world datasets often exhibit significant label noise. This paper examines the impact of such noise on object detection performance, a pivotal aspect of computer vision. We analyze the influence of noisy labels using three renowned object detection frameworks: YOLOv5, Faster R-CNN, and the recent YOLOv8, alongside established datasets: MS COCO, VOC, and ExDARK. Additionally, experiments with the UVM dataset explore domain-specific tasks in dense object scenarios. Two new metrics — Model Health and Detection Capability — were introduced to evaluate the results. Findings indicate that models maintain over 80% of their health (a 20% decline in mAP from the baseline) with up to 40% label corruption. However, Detection Capability deteriorates more sharply under the same conditions. The research also employs the D-RISE method for model explainability, highlighting crucial image regions affecting detection outcomes. Despite the noise, critical detection areas in models remain similar to those in clean data up to the 40% corruption level, as verified by similarity metrics. This study underscores the resilience of object detection models to label noise and provides insights into maintaining performance amidst data quality challenges.

A unified object and keypoint detection framework for Personal Protective Equipment use

Accurately detecting whether workers wear Personal Protective Equipment (PPE) in real time plays an important role in safety management. Previous studies mainly used multiple models jointly or only object detection for wearing relationship judgments. This makes it difficult to provide real-time, accurate detection of security relationships. Therefore, this paper proposes safe-wearing detection rules and a novel multi-targets and keypoints detection framework (MTKF), which is capable of accomplishing multiple classes of targets and keypoints detection simultaneously in one-stage, to get more accurate results. In order to improve the performance in the PPE and worker keypoints detection in challenging construction scenes, the detection head transformation strategy, mix group shuffle attention (MGSA) module, and the improved dual and cross-class suppression algorithm (DC-NMS) are proposed. The experimental results are implemented on one established dataset (Joint dataset) and two public datasets (SHWD and COCO), which conduct a comprehensive evaluation in multiple dimensions. Compared to the baseline model, our method improves the mAP by 2.6%–7.1%, reduces the number of parameters by at least 70%, and is able to achieve an inference speed of 155 fps.

Object Detection and Classification Framework for Analysis of Video Data Acquired from Indian Roads

Object detection and classification in autonomous vehicles are crucial for ensuring safe and efficient navigation through complex environments. This paper addresses the need for robust detection and classification algorithms tailored specifically for Indian roads, which present unique challenges such as diverse traffic patterns, erratic driving behaviors, and varied weather conditions. Despite significant progress in object detection and classification for autonomous vehicles, existing methods often struggle to generalize effectively to the conditions encountered on Indian roads. This paper proposes a novel approach utilizing the YOLOv8 deep learning model, designed to be lightweight, scalable, and efficient for real-time implementation using onboard cameras. Experimental evaluations were conducted using real-life scenarios encompassing diverse weather and traffic conditions. Videos captured in various environments were utilized to assess the model’s performance, with particular emphasis on its accuracy and precision across 35 distinct object classes. The experiments demonstrate a precision of 0.65 for the detection of multiple classes, indicating the model’s efficacy in handling a wide range of objects. Moreover, real-time testing revealed an average accuracy exceeding 70% across all scenarios, with a peak accuracy of 95% achieved in optimal conditions. The parameters considered in the evaluation process encompassed not only traditional metrics but also factors pertinent to Indian road conditions, such as low lighting, occlusions, and unpredictable traffic patterns. The proposed method exhibits superiority over existing approaches by offering a balanced trade-off between model complexity and performance. By leveraging the YOLOv8 architecture, this solution achieved high accuracy while minimizing computational resources, making it well suited for deployment in autonomous vehicles operating on Indian roads.

Night time Surveillance in Communication Using YOLOv3

Nighttime surveillance presents significant challenges due to low visibility, varying lighting conditions, and interference from artificial light sources. Hence, this study proposed an effective object detection framework using the YOLOv3 model to enhance real-time monitoring in night surveillance applications, specifically within communication and security systems. YOLOv3's architecture, with its multi-scale detection and use of predefined anchor boxes, enables robust detection of objects under low-light environments and amidst light interference from vehicles and streetlights. The proposed system is tested on a night surveillance dataset, where it demonstrates high precision and speed in identifying objects, making it suitable for real-time applications. With Mean Average Precision (mAP) 87.9%, YOLOv3 effectively balances detection accuracy with inference time, ensuring minimal latency in live surveillance feeds. The results indicate that YOLOv3 outperforms traditional models such as Faster RCNN, particularly in detecting small objects under poor illumination. This approach offers a reliable solution for enhancing communication and security systems in nighttime surveillance scenarios.

Consistency-based semi-supervised learning for oriented object detection

Semi-Supervised Object Detection (SSOD) has emerged as a potent framework that leverages unlabeled data to reduce annotation costs while enhancing model performance. Nevertheless, existing SSOD methods primarily concentrate on the conventional localization of horizontal objects, overlooking the common arbitrary-oriented objects in aerial scenes. Extending semi-supervised learning to oriented object detection still encounters two major challenges: (1) There is an inconsistency between classification and localization in oriented object detectors, which means that pseudo-labels selected solely based on classification scores cannot properly represent the localization quality. (2) The static and monotonic pseudo-label selection process fails to dynamically filter pseudo-labels for different tasks and categories throughout the iteration process, inevitably introducing accumulated noise and bias into the supervisory signals for the student model. To mitigate this problems, we proposed a novel consistency-based semi-supervised oriented object detection framework, named CSLO-Det. Specifically, Task Alignment Learning (TAL) is proposed to improve the consistency in terms of both features and learning objectives, thereby facilitating a more robust selection of pseudo-labels. Noise-Resistant Pseudo-label Mining (NR-PM) is introduced to separately and dynamically exploits positives for the classification and localization task. Moreover, diverging from the conventional semi-supervised learning frameworks that utilize pseudo-boxes, our method applies pixel-level dense supervision during unsupervised training, which improves detection performance. Comprehensive experiment results reveal that the proposed CSLO-Det attains state-of-the-art performance under multiple semi-supervised datasets.

Enhancing Inland Waterway Safety and Management through Machine Learning-Based Ship Detection

Efficient ship detection is essential for inland waterway management. Recent advances in artificial intelligence have prompted research in this field. This study introduces a real-time ship detection model utilizing computer vision and the YOLO object detection framework. The model is designed to identify and locate common inland waterway vessels, such as container ships, passenger vessels, barges, ferries, canoes, fishing boats, and sailboats. Data augmentation techniques were employed to enhance the model's ability to handle variations in ship appearance, weather, and image quality. The system achieved a mean Average Precision (mAP) of 98.4%, with precision and recall rates of 96.6% and 95.0%, respectively. These results demonstrate the model's effectiveness in practical applications. Its ability to generalize across diverse vessel types and environmental conditions suggests its potential integration into video surveillance for improved maritime safety, traffic control, and search and rescue operations.

Optimizing Object Detection Algorithms for Congenital Heart Diseases in Echocardiography: Exploring Bounding Box Sizes and Data Augmentation Techniques.

Congenital heart diseases (CHDs), particularly atrial and ventricular septal defects, pose significant health risks and common challenges in detection via echocardiography. Doctors often employ the cardiac structural information during the diagnostic process. However, prior CHD research has not determined the influence of including cardiac structural information during the labeling process and the application of data augmentation techniques. This study utilizes advanced artificial intelligence (AI)-driven object detection frameworks, specifically You Look Only Once (YOLO)v5, YOLOv7, and YOLOv9, to assess the impact of including cardiac structural information and data augmentation techniques on the identification of septal defects in echocardiographic images. The experimental results reveal that different labeling strategies substantially affect the performance of the detection models. Notably, adjustments in bounding box dimensions and the inclusion of cardiac structural details in the annotations are key factors influencing the accuracy of the model. The application of deep learning techniques in echocardiography enhances the precision of detecting septal heart defects. This study confirms that careful annotation of imaging data is crucial for optimizing the performance of object detection algorithms in medical imaging. These findings suggest potential pathways for refining AI applications in diagnostic cardiology studies.

A novel multiscale cGAN approach for enhanced salient object detection in single haze images

In computer vision, image dehazing is a low-level task that employs algorithms to analyze and remove haze from images, resulting in haze-free visuals. The aim of Salient Object Detection (SOD) is to locate the most visually prominent areas in images. However, most SOD techniques applied to visible images struggle in complex scenarios characterized by similarities between the foreground and background, cluttered backgrounds, adverse weather conditions, and low lighting. Identifying objects in hazy images is challenging due to the degradation of visibility caused by atmospheric conditions, leading to diminished visibility and reduced contrast. This paper introduces an innovative approach called Dehaze-SOD, a unique integrated model that addresses two vital tasks: dehazing and salient object detection. The key novelty of Dehaze-SOD lies in its dual functionality, seamlessly integrating dehazing and salient object identification into a unified framework. This is achieved using a conditional Generative Adversarial Network (cGAN) comprising two distinct subnetworks: one for image dehazing and another for salient object detection. The first module, designed with residual blocks, Dark Channel Prior (DCP), total variation, and the multiscale Retinex algorithm, processes the input hazy images. The second module employs an enhanced EfficientNet architecture with added attention mechanisms and pixel-wise refinement to further improve the dehazing process. The outputs from these subnetworks are combined to produce dehazed images, which are then fed into our proposed encoder–decoder framework for salient object detection. The cGAN is trained with two modules working together: the generator aims to produce haze-free images, whereas the discriminator distinguishes between the generated haze-free images and real haze-free images. Dehaze-SOD demonstrates superior performance compared to state-of-the-art dehazing methods in terms of color fidelity, visibility enhancement, and haze removal. The proposed method effectively produces high-quality, haze-free images from various hazy inputs and accurately detects salient objects within them. This makes Dehaze-SOD a promising tool for improving salient object detection in challenging hazy conditions. The effectiveness of our approach has been validated using benchmark evaluation metrics such as mean absolute error (MAE), peak signal-to-noise ratio (PSNR), and structural similarity index measure (SSIM).

ADWNet: An improved detector based on YOLOv8 for application in adverse weather for autonomous driving

AbstractDrawing inspiration from the state‐of‐the‐art object detection framework YOLOv8, a new model termed adverse weather net (ADWNet) is proposed. To enhance the model's feature extraction capabilities, the efficient multi‐scale attention (EMA) module has been integrated into the backbone. To address the problem of information loss in fused features, Neck has been replaced with RepGDNeck. Simultaneously, to expedite the model's convergence, the bounding box's loss function has been optimized to SIoU loss. To elucidate the advantages of ADWNet in the context of adverse weather conditions, ablation studies and comparative experiments were conducted. The results indicate that although the model's parameter count increased by 18.4%, the accuracy for detecting rain, snow, and fog in adverse weather conditions improved by 22%, while the FLOPs (floating point operations) decreased by 5%. The results of the comparison experiments conducted on the WEDGE dataset show that ADWNet outperforms other object detection models in adverse weather in terms of accuracy, model parameters and FLOPs. To validate ADWNet's real‐world efficacy, data was extracted from a car recorder under adverse conditions on highways, visual inference was conducted, and its accuracy was demonstrated in interpreting real‐world scenarios. The config files are available at https://github.com/Xinyun‐Feng/ADWNet.

DBTSF-VSOD: a decision-based two-stage framework for video salient object detection

DBTSF-VSOD: a decision-based two-stage framework for video salient object detection

Optimized automated blood cells analysis using Enhanced Greywolf Optimization with integrated attention mechanism and YOLOv5

Blood testing is widely regarded as a fundamental diagnostic procedure in healthcare, with blood cell counting being particularly vital for diagnosing conditions and evaluating overall health. Traditional methods employing hemocytometers and diverse laboratory instruments are typically employed to conduct blood counts manually. Using a microscope, manually counting, and examining blood cells takes time and effort. Thus, developing autonomous blood cells detecting and counting cells in this system becomes necessary to help doctors diagnose patients quickly and accurately. The proposed method integrates the YOLOv5 object detection framework with Enhanced Graywolf Optimization (EGWO) and an Attention Mechanism to improve the accuracy and efficiency of RBC detection. The YOLOv5 architecture is optimized for the task through the EGWO algorithm, which fine-tunes the model parameters, while the Attention Mechanism focuses on extracting critical features from blood cell images. The model is trained and validated using the BCCD dataset, employing an 80/20 split for training and validation. Data augmentation techniques, such as random rotations, flips, and color adjustments, are applied to enhance model generalization. Experimental results demonstrate that our method achieves a high accuracy of 99.89 %, outperforming existing models like Faster R-CNN, CNN, and DCNN regarding accuracy, training time, and inference speed. The proposed method's moderate complexity and fast inference time make it suitable for real-time applications in clinical settings. This research provides a robust and efficient solution for automated RBC detection and counting, with potential implications for improving diagnostic processes in hematological analyses. Overall, in practical applications, it is helpful for counting red blood cells from smeared pictures in less than a second.

Drone-Based Visible–Thermal Object Detection with Transformers and Prompt Tuning

The use of unmanned aerial vehicles (UAVs) for visible–thermal object detection has emerged as a powerful technique to improve accuracy and resilience in challenging contexts, including dim lighting and severe weather conditions. However, most existing research relies on Convolutional Neural Network (CNN) frameworks, limiting the application of the Transformer’s attention mechanism to mere fusion modules and neglecting its potential for comprehensive global feature modeling. In response to this limitation, this study introduces an innovative dual-modal object detection framework called Visual Prompt multi-modal Detection (VIP-Det) that harnesses the Transformer architecture as the primary feature extractor and integrates vision prompts for refined feature fusion. Our approach begins with the training of a single-modal baseline model to solidify robust model representations, which is then refined through fine-tuning that incorporates additional modal data and prompts. Tests on the DroneVehicle dataset show that our algorithm achieves remarkable accuracy, outperforming comparable Transformer-based methods. These findings indicate that our proposed methodology marks a significant advancement in the realm of UAV-based object detection, holding significant promise for enhancing autonomous surveillance and monitoring capabilities in varied and challenging environments.

Performance evaluation of semi-supervised learning frameworks for multi-class weed detection.

Precision weed management (PWM), driven by machine vision and deep learning (DL) advancements, not only enhances agricultural product quality and optimizes crop yield but also provides a sustainable alternative to herbicide use. However, existing DL-based algorithms on weed detection are mainly developed based on supervised learning approaches, typically demanding large-scale datasets with manual-labeled annotations, which can be time-consuming and labor-intensive. As such, label-efficient learning methods, especially semi-supervised learning, have gained increased attention in the broader domain of computer vision and have demonstrated promising performance. These methods aim to utilize a small number of labeled data samples along with a great number of unlabeled samples to develop high-performing models comparable to the supervised learning counterpart trained on a large amount of labeled data samples. In this study, we assess the effectiveness of a semi-supervised learning framework for multi-class weed detection, employing two well-known object detection frameworks, namely FCOS (Fully Convolutional One-Stage Object Detection) and Faster-RCNN (Faster Region-based Convolutional Networks). Specifically, we evaluate a generalized student-teacher framework with an improved pseudo-label generation module to produce reliable pseudo-labels for the unlabeled data. To enhance generalization, an ensemble student network is employed to facilitate the training process. Experimental results show that the proposed approach is able to achieve approximately 76% and 96% detection accuracy as the supervised methods with only 10% of labeled data in CottonWeedDet3 and CottonWeedDet12, respectively. We offer access to the source code (https://github.com/JiajiaLi04/SemiWeeds), contributing a valuable resource for ongoing semi-supervised learning research in weed detection and beyond.

A streamlined framework for BEV-based 3D object detection with prior masking

In the field of autonomous driving, perception tasks based on Bird's-Eye-View (BEV) have attracted considerable research attention due to their numerous benefits. Despite recent advancements in performance, efficiency remains a challenge for real-world implementation. In this study, we propose an efficient and effective framework that constructs a spatio-temporal BEV feature from multi-camera inputs and leverages it for 3D object detection. Specifically, the success of our network is primarily attributed to the design of the lifting strategy and a tailored BEV encoder. The lifting strategy is tasked with the conversion of 2D features into 3D representations. In the absence of depth information in the images, we innovatively introduce a prior mask for the BEV feature, which can assess the significance of the feature along the camera ray at a low cost. Moreover, we design a lightweight BEV encoder, which significantly boosts the capacity of this physical-interpretation representation. In the encoder, we investigate the spatial relationships of the BEV feature and retain rich residual information from upstream. To further enhance performance, we establish a 2D object detection auxiliary head to delve into insights offered by 2D object detection and leverage the 4D information to explore the cues within the sequence. Benefiting from all these designs, our network can capture abundant semantic information from 3D scenes and strikes a balanced trade-off between efficiency and performance.