Bounding Box Research Articles

How Does Image Complexity Affect the Accuracy of an Interactive Image Segmentation Algorithm?

This study investigates the impact of image complexity on the accuracy of interactive image segmentation algorithms. Image complexity plays a crucial role in segmentation performance, yet previous studies have primarily relied on subjective methods, leaving a gap in understanding how objective measures impact accuracy. The purpose of this research is to explore the relationship between image complexity and segmentation performance and to propose an adaptive approach for improving accuracy based on complexity measures. The study utilizes objective measures, namely entropy and fractal dimension, to quantify image complexity. An interactive image segmentation algorithm is employed, with a bounding box as the background and strokes as the foreground annotations. The number of strokes is dynamically adjusted based on complexity measures, ensuring a tailored segmentation approach. Comparative evaluations are conducted to assess the effectiveness of dynamic and fixed stroke allocation strategies. The principal results reveal a significant influence of image complexity on segmentation accuracy. The dynamic stroke allocation strategy outperforms fixed allocation, highlighting the importance of adapting to complexity. Moreover, the optimal combination of strokes and superpixel sizes is explored, providing valuable insights for practitioners. The findings emphasize the need to consider image complexity in segmentation algorithms to achieve accurate results. In conclusion, this study contributes to the understanding of the relationship between image complexity and interactive image segmentation. By introducing a dynamic stroke allocation approach and evaluating different configurations, the research provides insights into optimizing accuracy based on image complexity. The adaptive strategy improves segmentation performance and guides the development of robust algorithms. Future research can further refine the adaptive approach, explore additional complexity measures, and incorporate advanced machine learning techniques to enhance interactive image segmentation. Overall, this study advances the field by highlighting the importance of image complexity, providing guidance for practitioners, and paving the way for more efficient segmentation algorithms.

A high-quality trajectory generation method for the multi-person tracking

Recently, Multi-Person Tracking (MPT) has experienced significant progress due to the emergence of innovative methods. However, congestion and occlusion persist as formidable challenges for MPT. In this study, we propose a Space Matching method, which systematically considers all the potential pedestrian detections within each spatial division. This method can effectively mitigate the erroneous matching problem and hence improve the matching performance. We enhance the update process of the Kalman filter by excluding distorted detections, which can realize a more accurate motion estimation. Further, we propose a Locally Weighted Interpolation method to reconstruct the missing bounding boxes on certain frames, which can significantly reduce the computation time, saving about 80% compared to the Gaussian process regression-based approach. With the help of these strategies, we can generate high-quality pedestrian trajectories, ultimately bringing state-of-the-art results on both MOT17 and MOT20 datasets.

Advancements in Image Enhancement and Attention based EfficientDet Optimization Classifier for Precise Osteosarcoma Lung Nodule Detection

Introduction: Osteosarcoma is a malignant bone tumor that frequently spreads to the lungs, hence therapy effectiveness depends on early identification. However, noise and subtle characteristics still pose a challenge for reliable Lung Nodules Detection (LND) in medical pictures. In earlier work, SSD-VGG16 was implemented to provide a bounding box with an accuracy score that represented a single osteosarcoma nodule. Increasing model complexity is sometimes necessary to achieve improved accuracy with current approaches, which might worsen their computing inefficiencies.Methods: For accurate osteosarcoma lung nodule identification, this study offers the hybrid Dynamic Virtual Bats Algorithm with Attention based Efficient Object identification (A- EfficientDet). In order to improve the quality and informativeness of clinical pictures, this study suggests including Chebyshev filtering into the pre-processing pipeline. It focuses on CT scans for the purpose of detecting lung nodules associated with osteosarcoma. Additionally, provide the optimized A-EfficientDet model, a hybrid EfficientDet model improved using the DVBA optimization technique for accurate lung nodule identification. Results: The effectiveness of the suggested strategy in attaining accurate osteosarcoma LND is demonstrated by the experimental findings. Chebyshev filtering is incorporated during the pre-processing step, which leads to more accurate detection findings by improving the signal-to-noise ratio (SNR) and lung nodule visibility. Conclusion: Additionally, the improved EfficientDet model demonstrates its suitability for clinical applications in early osteosarcoma detection and treatment monitoring by achieving (SOTA) State-Of-The-Art execution by the metrics of sensitivity, specificity, and F1 score.

An improved LIO-SAM algorithm by integrating image information for dynamic and unstructured environments

Abstract Simultaneous localization and mapping (SLAM) is the process of estimating the trajectory of a mobile sensor carrier and creating a representation of its surroundings. Traditional SLAM algorithms are based on "static world assumption" and simplify the problem by filtering out moving objects or tracking them separately in complex dynamic environments. However, this strong assumption restricts the application of SLAM algorithms on highly dynamic and unstructured environments. In order to resolve above problem, this paper propose an improved object-aware dynamic SLAM system by integrating image information, i.e., semantic and velocity information. Firstly, we adopt deep learning method to detect both the 2D and 3D bounding boxes of objects in the environment. This information is then used to perform multi-view, multi-dimensional bundle optimization to jointly refine the poses of camera, object, and point. Secondly, 2D detection results from image and 3D detection results from lidar are integrated by the joint probabilistic data association (JPDA) data association algorithm to facilitate object-level data association. We also calculate 2D and 3D motion velocity and this information is used to constraint the motion of the object. Finally, we perform comprehensive experiments on different datasets, including NCLT, M2DGR, and KITTI to prove the performance of the proposed method.

Evaluation of a Semi-Automated Wound-Halving Algorithm for Split-Wound Design Studies: A Step towards Enhanced Wound-Healing Assessment.

Background: Chronic leg ulcers present a global challenge in healthcare, necessitating precise wound measurement for effective treatment evaluation. This study is the first to validate the "split-wound design" approach for wound studies using objective measures. We further improved this relatively new approach and combined it with a semi-automated wound measurement algorithm. Method: The algorithm is capable of plotting an objective halving line that is calculated by splitting the bounding box of the wound surface along the longest side. To evaluate this algorithm, we compared the accuracy of the subjective wound halving of manual operators of different backgrounds with the algorithm-generated halving line and the ground truth, in two separate rounds. Results: The median absolute deviation (MAD) from the ground truth of the manual wound halving was 2% and 3% in the first and second round, respectively. On the other hand, the algorithm-generated halving line showed a significantly lower deviation from the ground truth (MAD = 0.3%, p < 0.001). Conclusions: The data suggest that this wound-halving algorithm is suitable and reliable for conducting wound studies. This innovative combination of a semi-automated algorithm paired with a unique study design offers several advantages, including reduced patient recruitment needs, accelerated study planning, and cost savings, thereby expediting evidence generation in the field of wound care. Our findings highlight a promising path forward for improving wound research and clinical practice.

Discriminative action tubelet detector for weakly-supervised action detection

We propose a novel framework for spatiotemporal action detection using only video-level class labels as weak supervision. Traditional fully-supervised approaches rely on labor-intensive manual annotation of bounding boxes for each frame. In contrast, collecting video-level class labels is significantly less tedious and more feasible compared to annotating frame-level sequences with bounding boxes. To address this challenge, we propose a discriminative action tubelet detector, called DAT-detector, designed to discern discriminative tubelets from action tubelet proposals (ATPs). Whereas the previous approaches have only focused on tubelet selection among the predefined object proposals, our DAT-detector prioritizes the generation of more precise action tubelets using regression and attention modules. Moreover, we introduce an ATP generation method that enhances the quality of tubelet proposals. Our approach achieves state-of-the-art performance on several benchmarks, and also demonstrates competitive performance even with fully-supervised approaches.

Localizing unknown nodes with an FPGA-enhanced edge computing UAV in wireless sensor networks: Implementation and evaluation

Great interest is directed toward real-time applications to determine the exact location of sensor nodes deployed in an area of interest. In this paper, we present a novel approach using a combination of the Kalman filter and regularized bounding box method for localizing unknown nodes in an area using an FPGA-enhanced edge computing UAV whose trajectory is known and is represented as the position of many anchors. The UAV is equipped with a GPS system that allows it to gather location data of sensor nodes as it moves around its environment. We employ a regularized bounding box to predict the positions of the unknown nodes using regularization factors and we use the Kalman filter algorithm to smooth and improve the accuracy of the sensor nodes to be localized. In order to localize the unknown nodes, the UAV receives the number of hops from each node and uses this information as input to the localization algorithm. Furthermore, the use of an FPGA board allows for real-time processing of sensory data, enabling the UAV to make fast and accurate decisions in dynamic environments. The localization algorithm was implemented on the FPGA board “Zynq MiniZed 7007s evaluation board” using Xilinx blocks in Simulink, and the generated code was converted into VHDL using Xilinx System Generator. The algorithm was simulated and synthesized using “Vivado” software. In fact, the proposed system was evaluated by comparing the performances achieved through two different implementations: Hardware and Software implementation. In effect, the performance of FPGA hardware implementation presents a new achievement in localization due to its easy testing and fast implementation. Our results show that this approach can efficiently locate unknown nodes with good latency and high accuracy. In fact, the execution time of the FPGA-integrated algorithm is reduced by about 60 times compared to the software implementation and the power consumption is about 100 mW, which proves the suitability of FPGA for localization in WSNs, offering a promising solution for various mobile WSN applications.

An Improved YOLOv8 Model for Lotus Seedpod Instance Segmentation in the Lotus Pond Environment

Lotus seedpod maturity detection and segmentation in pond environments play a significant role in yield prediction and picking pose estimation for lotus seedpods. However, it is a great challenge to accurately detect and segment lotus seedpods due to insignificant phenotypic differences between the adjacent maturity, changing illumination, overlap, and occlusion of lotus seedpods. The existing research pays attention to lotus seedpod detection while ignoring maturity detection and segmentation problems. Therefore, a semantic segmentation dataset of lotus seedpods was created, where a copy-and-paste data augmentation tool was employed to eliminate the class-imbalanced problem and improve model generalization ability. Afterwards, an improved YOLOv8-seg model was proposed to detect and segment the maturity of lotus seedpods. In the model, the convolutional block attention module (CBAM) was embedded in the neck network to extract distinguished features of different maturity stages with negligible computation cost. Wise-Intersection over Union (WIoU) regression loss function was adopted to refine the regression inference bias and improve the bounding box prediction accuracy. The experimental results showed that the proposed YOLOv8-seg model provides an effective method for “ripe” and “overripe” lotus seedpod detection and instance segmentation, where the mean average precision of segmentation mask (mAPmask) reaches 97.4% and 98.6%, respectively. In addition, the improved YOLOv8-seg exhibits high robustness and adaptability to complex illumination in a challenging environment. Comparative experiments were conducted using the proposed YOLOv8-seg and other state-of-the-art instance segmentation methods. The results showed that the improved model is superior to the Mask R-CNN and YOLACT models, with recall, precision, mAPbox and mAPmask being 96.5%, 94.3%, 97.8%, and 98%, respectively. The average running time and weight size of the proposed model are 25.9 ms and 7.4 M, respectively. The proposed model obtained the highest mAP for lotus seedpod maturity detection and segmentation while maintaining an appropriate model size and speed. Furthermore, based on the obtained segmentation model, 3D visualization of the lotus pond scene is performed, and cloud point of lotus seedpods is generated, which provides a theoretical foundation for robot harvesting in the lotus pond.

CRGF-YOLO: An Optimized Multi-Scale Feature Fusion Model Based on YOLOv5 for Detection of Steel Surface Defects

The identification of imperfections on steel surfaces is vital for ensuring the quality of industrial products. It requires the capability of real-time detection with high accuracy. This paper proposes the CRGF-YOLO (Contextual Reparameterized Generalized Feature) model based on YOLOv5. In the network, BottleneckCSP structures and depthwise separable convolutions utilizing the structural reparameterization are introduced to reduce the model size and improve performance. In addition, contextual transformer modules are employed as self-attention mechanisms to improve feature representations by capturing long-range dependencies, outperforming conventional convolutional networks. Furthermore, the simplified generalized feature pyramid network is embedded to aggregate multi-scale feature maps and enhance the network’s robustness. Finally, four prediction heads with different sizes are employed to predict defects, which are supported by prior bounding boxes generated using k-means clustering algorithm. The Focal-EIOU (Exponential Intersection over Union) loss function is introduced to improve detection accuracy and expedite model convergence. The improved model achieves a mean average precision (mAP) of 82.2% on the NEU-DET dataset, outperforming the baseline YOLOv5s by 7.7% mAP while maintaining real-time speeds. Comparative evaluations demonstrate CRGF-YOLO’s superior performance over previous state-of-the-art methods like Faster R-CNN (77.4% mAP), YOLOv3 (77.4% mAP), YOLOv7s (72.1% mAP), and YOLOv8s (78.7% mAP) for steel surface defect detection. Overall, this study provides valuable insights and practical guidance for the advancement of defect detection technology.

Head motion-corrected eye gaze tracking with the da Vinci surgical system.

To facilitate the integration of point of gaze (POG) as an input modality for robot-assisted surgery, we introduce a robust head movement compensation gaze tracking system for the da Vinci Surgical System. Previous surgical eye gaze trackers require multiple recalibrations and suffer from accuracy loss when users move from the calibrated position. We investigate whether eye corner detection can reduce gaze estimation error in a robotic surgery context. A polynomial regressor is first used to estimate POG after an 8-point calibration, and then, using another regressor, the POG error from head movement is estimated from the shift in 2D eye corner location. Eye corners are computed by first detecting regions of interest using the You Only Look Once (YOLO) object detector trained on 1600 annotated eye images (open dataset included). Contours are then extracted from the bounding boxes and a derivative-based curvature detector refines the eye corner. Through a user study (n = 24), our corner-contingent head compensation algorithm showed an error reduction in degrees of visual angle of 1.20 (p = 0.037) for the left eye and 1.26 (p = 0.079) for the right compared to the previous gold-standard POG error correction method. In addition, the eye corner pipeline showed a root-mean-squared error of 3.57 (SD = 1.92) pixels in detecting eye corners over 201 annotated frames. We introduce an effective method of using eye corners to correct for eye gaze estimation, enabling the practical acquisition of POG in robotic surgery.

Modified M‐RCNN approach for abandoned object detection in public places

AbstractDetection of abandoned and stationary objects like luggage, boxes, machinery, and so forth, in public places is one of the challenging and critical tasks in the video surveillance system. These objects may contain weapons, bombs, or other explosive materials that threaten the public. Though various applications have been developed to detect stationary objects, different challenges, like occlusions, changes in geometrical features of things, and so forth, are still to be addressed. Considering the complexity of scenarios in public places and the variety of objects, a context‐aware model is developed based on mask region‐based convolution network (M‐RCNN) for detecting abandoned objects. A modified convolution operation is implemented in the Backbone network to understand features from geometric variations near objects. These modified operation layers can be adapted based on geometric interpretations to extract required features. Finally, a bounding box operation is performed to locate the abandoned object and mask the particular thing. Experiments have been performed on the benchmark dataset like ABODA and our dataset, which shows that an mAP of 0. 0.699 is achieved for model 1, 0.675 is achieved for model 2, and 0.734 mAP is completed for model 3. An ablation analysis has also been performed and compared with other state‐of‐the‐art methods. Based on the results, the proposed model better detects abandoned objects than existing state‐of‐the‐art methods.

SRFAD-Net: Scale-Robust Feature Aggregation and Diffusion Network for Object Detection in Remote Sensing Images

The significant differences in target scales of remote sensing images lead to remarkable variations in visual features, posing significant challenges for feature extraction, fusion, regression, and classification. For example, models frequently struggle to capture features of targets across all scales, inadequately consider the weights and importance of features at different scales during fusion, and encounter accuracy limitations when detecting targets of varying scales. To tackle these challenges, we proposes a Scale-Robust Feature Aggregation and Diffusion Network (SRFAD-Net) for remote sensing target detection. This model includes a Scale-Robust Feature Network (SRFN), an Adaptive Feature Aggregation and Diffusion (AFAD) module, and a Focaler-GIoU Loss. SRFN extracts scale-robust features by constructing a multi-scale pyramid. It includes a downsampling (ADown) module that combines the advantages of average pooling and max pooling, effectively preserving background information and salient features. This further enhances the network’s ability to handle targets of varying scales and shapes. The introduced Deformable Attention(DAttention) mechanism captures target features effectively through adaptive adjustment of the receptive field’s shape and size, reducing background clutter and substantially enhancing the model’s performance in detecting distant objects. In the feature fusion stage, we propose the AFAD module, which utilizes a dimension-adaptive perceptual selection mechanism and parallel depthwise convolutional operations to precisely aggregate multi-channel information. It then employs a diffusion mechanism to spread contextual information across various scales, greatly improving the network’s ability to extract and fuse features across multiple scales. For the detection head, we adopt the Focaler-GIoU Loss, leveraging its advantages in handling non-overlapping bounding boxes, effectively alleviating the difficulty of localization caused by scale variations. We have undertaken experiments on two widely utilized aerial target datasets: the Remote Sensing Scene Object Detection Dataset (RSOD) and NWPU VHR-10, which is a high-resolution object detection dataset from Northwestern Polytechnical University. The findings of these experiments clearly illustrate that SRFAD-Net surpasses the performances of mainstream detectors.

Infrared Dim Small Target Detection Networks: A Review.

In recent years, with the rapid development of deep learning and its outstanding capabilities in target detection, innovative methods have been introduced for infrared dim small target detection. This review comprehensively summarizes public datasets, the latest networks, and evaluation metrics for infrared dim small target detection. This review mainly focuses on deep learning methods from the past three years and categorizes them based on the six key issues in this field: (1) enhancing the representation capability of small targets; (2) improving the accuracy of bounding box regression; (3) resolving the issue of target information loss in the deep network; (4) balancing missed detections and false alarms; (5) adapting for complex backgrounds; (6) lightweight design and deployment issues of the network. Additionally, this review summarizes twelve public datasets for infrared dim small targets and evaluation metrics used for detection and quantitatively compares the performance of the latest networks. Finally, this review provides insights into the future directions of this field. In conclusion, this review aims to assist researchers in gaining a comprehensive understanding of the latest developments in infrared dim small target detection networks.

An adaptive detection approach for multi-scale defects on wind turbine blade surface

Multi-scale defects will inevitably appear in wind turbine blade defect detection within practical detection applications. An adaptive multi-scale detection approach is proposed to accurately classify and locate defects on the wind turbine blade surface. The proposed approach includes two main detection procedures: rough detection and precise detection. In the rough detection, the multi-level features extraction module with an adaptive bounding box proposal module is used to depict multi-scale defect regions and train a binary classifier to distinguish defects from non-defects. At the precise detection stage, three defect categories and four coordinates representing defect locations are obtained based on a multi-class defect classifier and regression of rough location boxes. The proposed method is evaluated on a real wind turbine blade surface defect dataset collected in a commercial wind farm and annotated manually. Results show that (1) the proposed model can detect the class of the blade multi-scale defects and outperforms other schemes with 96.89% mAP in the same model training epochs, (2) the positioning performance analysis of the model for multi-scale defects is conducted to validate the accuracy of the proposed model for multi-scale defect location.

A novel robotic-assisted deep learning-enabled computer vision approach for nondestructive diagnosis of railway bolt faults

Abstract Railways play a vital role in the inland transportation system worldwide, and abnormal bolt components at the track joints are the main cause of train accidents. The detection and identification of faults in rail bolt components are of considerable research importance. To address this problem, numerous researchers have opted for computer vision-based methods to accomplish the detection and identification of the target, but the existing methods have poor detection performance diminished detection capabilities when the target position changes or some feature information is occluded, and the detection speed and accuracy are far from meeting the requirements of practical applications. Therefore, based on the construction of a dedicated dataset for bolt components, this paper uses the K-means dimensional clustering algorithm to re-cluster the dataset according to the target size characteristics, with the aim of reduce the bounding box regression error. At the same time, a novel loss function iteration method is proposed by incorporating an adaptive optimization algorithm, in order to improve the detection speed and ensure good convergence, and the model complexity is reduced based on deep model pruning. Finally, the optimized detection model is implemented on the robotic-assisted platform for testing, and the experimental results verify that the algorithm can quickly and accurately complete various fault diagnosis tasks of bolt components in practical applications. The main achievements of this study include the construction of a large-scale image dataset for novel rail bolt components and propelled the application of deep learning methods in vision-based rail bolt fault diagnosis.

UniQRNet: Unifying Referring Expression Grounding and Segmentation with QRNet

Referring expression comprehension aims to align natural language queries with visual scenes, which requires establishing fine-grained correspondence between vision and language. This has important applications in multi-modal reasoning systems. Existing methods typically use text-agnostic visual backbones to extract features independently without considering the specific text input. However, we argue that the extracted visual features can be inconsistent with the referring expression, which hurts multi-modal understanding. To address this, we first propose Query-modulated Refinement Network (QRNet) that leverages language guidance to guide visual feature extraction. However, it only focuses on the grounding task that can only provide coarse-grained annotations in the form of bounding box coordinates. The guidance for the visual backbone is indirect, and the inconsistent issue still exists. To this end, we further propose UniQRNet, a multi-task framework over the QRNet to learn referring expression grounding and segmentation jointly. The framework introduces a multi-task head that leverages fine-grained pixel-level supervision from the segmentation task to directly guide the intermediate layers of QRNet to learn text-consistent visual features. Besides, UniQRNet also includes a loss balance strategy that allows two types of supervision signals to cooperate and optimize the model together. We conduct the most comprehensive comparison experiment covering four major datasets, ten evaluation set and three evaluation metrics used in previous work. UniQRNet outperforms previous state-of-the-art methods by a large margin on both referring comprehensive grounding (1.8%~5.09%) and segmentation tasks (0.57%~5.56%). Ablation and analysis reveal that UniQRNet can improve the consistency of visual features with text input and can bring significant performance improvement.

Use of Yolo Detection for 3D Pose Tracking of Cardiac Catheters Using Bi-Plane Fluoroscopy

The increasing rate of minimally invasive procedures and the growing prevalence of cardiovascular disease have led to a demand for higher-quality guidance systems for catheter tracking. Traditional methods for catheter tracking, such as detection based on single points and applying masking techniques, have been limited in their ability to provide accurate pose information. In this paper, we propose a novel deep learning-based method for catheter tracking and pose detection. Our method uses a Yolov5 bounding box neural network with postprocessing to perform landmark detection in four regions of the catheter: the tip, radio-opaque marker, bend, and entry point. This allows us to track the catheter’s position and orientation in real time, without the need for additional masking or segmentation techniques. We evaluated our method on a dataset of fluoroscopic images from two distinct datasets and achieved state-of-the-art results in terms of accuracy and robustness. Our model was able to detect all four landmark features (tip, marker, bend, and entry) used to generate a pose for a catheter with 0.285 ± 0.143 mm, 0.261 ± 0.138 mm, 0.424 ± 0.361 mm, and 0.235 ± 0.085 mm accuracy. We believe that our method has the potential to significantly improve the accuracy and efficiency of catheter tracking in medical procedures that utilize bi-plane fluoroscopy guidance.

Swin-Roleaf: A new method for characterizing leaf azimuth angle in large-scale maize plants

Maize is the most productive grain crop in the world in terms of total production and yield. Leaf azimuth angle, the directional orientation of the maize leaf, is an important characteristic that affects the efficiency of sunlight reception and photosynthesis in maize canopies. To study maize plant architecture, most of the previous studies have been focusing on leaf vertical angle characterization due to the inherent difficulty of measuring leaf azimuth angle accurately. Achieving accurate and efficient measurement of leaf azimuth angle in maize is important for yield improvement. Traditionally, leaf azimuth angle is measured manually using a protractor or manual annotations of images, which is time-consuming, costly, and prone to human error. Recent advances in Unmanned Aerial Vehicle (UAV) and artificial intelligence have shown great potential in leaf azimuth angle detection. However, most of the deep learning algorithms for object detection are developed for Horizontal Bounding Boxes (HBBs). To measure the azimuth angle of maize leaves, a robust Oriented Bounding Boxes (OBBs) detection algorithm is needed. In this study, a dataset containing top-view images of maize plants was collected using an UAV. An automated tool, Swin-Roleaf, was developed for maize leaf detection. It is an improvement of the YOLOv5 model, uniquely integrating the Swin Transformer (ST) and Circular Smooth Label (CSL). Swin-Roleaf was able to accurately detect OBBs of individual leaves and thereby enable the estimation of leaf azimuth angle. The experimental results showed that Swin-Roleaf was able to accurately detect maize leaves and measure the azimuth angle with a model OBBAP50 of 90.59% on OBBs. The system-derived leaf azimuth angle was highly correlated with the ground truth with a correlation coefficient larger than 0.96. Our study demonstrates for the first time the feasibility of using UAVs coupled with deep learning algorithms to detect leaf azimuth angles under field conditions. The system can significantly speed up the measurement of maize leaf azimuth angles and provide a new tool to accelerate breeding programs.

Linear Gaussian bounding box representation and ring-shaped rotated convolution for oriented object detection

In oriented object detection, current representations of oriented bounding boxes (OBBs) often suffer from the boundary discontinuity problem. Methods of designing continuous regression losses do not essentially solve this problem. Although Gaussian bounding box (GBB) representation avoids this problem, directly regressing GBB is susceptible to numerical instability. We propose linear GBB (LGBB), a novel OBB representation. By linearly transforming the elements of GBB, LGBB avoids the boundary discontinuity problem and has high numerical stability. In addition, existing convolution-based rotation-sensitive feature extraction methods only have local receptive fields, resulting in slow feature aggregation. We propose ring-shaped rotated convolution (RRC), which adaptively rotates feature maps to arbitrary orientations to extract rotation-sensitive features under a ring-shaped receptive field, rapidly aggregating features and contextual information. Experimental results demonstrate that LGBB and RRC achieve state-of-the-art performance. Furthermore, integrating LGBB and RRC into various models effectively improves detection accuracy.

HR-YOLO: A Multi-Branch Network Model for Helmet Detection Combined with High-Resolution Network and YOLOv5

Automatic detection of safety helmet wearing is significant in ensuring safe production. However, the accuracy of safety helmet detection can be challenged by various factors, such as complex environments, poor lighting conditions and small-sized targets. This paper presents a novel and efficient deep learning framework named High-Resolution You Only Look Once (HR-YOLO) for safety helmet wearing detection. The proposed framework synthesizes safety helmet wearing information from the features of helmet objects and human pose. HR-YOLO can use features from two branches to make the bounding box of suppression predictions more accurate for small targets. Then, to further improve the iterative efficiency and accuracy of the model, we design an optimized residual network structure by using Optimized Powered Stochastic Gradient Descent (OP-SGD). Moreover, a Laplace-Aware Attention Model (LAAM) is designed to make the YOLOv5 decoder pay more attention to the feature information from human pose and suppress interference from irrelevant features, which enhances network representation. Finally, non-maximum suppression voting (PA-NMS voting) is proposed to improve detection accuracy for occluded targets, using pose information to constrain the confidence of bounding boxes and select optimal bounding boxes through a modified voting process. Experimental results demonstrate that the presented safety helmet detection network outperforms other approaches and has practical value in application scenarios. Compared with the other algorithms, the proposed algorithm improves the precision, recall and mAP by 7.27%, 5.46% and 7.3%, on average, respectively.