ATDIOU: Arctangent Differential Loss Function for Bounding Box Regression.

Object detection remains a cornerstone task in computer vision, with wide-ranging applications in autonomous driving, surveillance, and medical imaging. However, traditional methods rely heavily on large annotated datasets, limiting their adaptability in low-resource environments. We propose Hybrid Masked Autoencoder-Detection Transformer++ (HybridMAE-DETR++), a novel self-supervised object detection framework that synergizes Masked Autoencoders (MAEs) and DINOv2 for Vision Transformer (ViT) pretraining. Integrated with a Swin-ViT hybrid backbone and an enhanced DETR++ detection head, the framework significantly reduces dependence on annotated data while improving detection accuracy for small and occluded objects. Evaluated on COCO 2017 and Cityscapes, HybridMAE-DETR++ achieves 47.5% mean Average Precision (mAP) and 68.0% Intersection over Union (IoU) on COCO, and 53.2% mAP and 72.1% IoU on Cityscapes, outperforming DETR and other transformer-based baselines. Ablation and sensitivity analyses confirm the robustness of our hybrid pretraining strategy, and visualizations using Layer-weighted Class Activation Mapping (LayerCAM) and Gradient-weighted Class Activation Mapping++ (Grad-CAM++) validate model interpretability. Despite a moderate increase in training time, the precision gains justify the computational cost. This framework sets a new benchmark for label-efficient, interpretable object detection in real-world scenarios.

Intersection over Union (IoU) is the most popular evaluation metric used in the object detection benchmarks. However, there is a gap between optimizing the commonly used distance losses for regressing the parameters of a bounding box and maximizing this metric value. The optimal objective for a metric is the metric itself. In the case of axis-aligned 2D bounding boxes, it can be shown that $IoU$ can be directly used as a regression loss. However, $IoU$ has a plateau making it infeasible to optimize in the case of non-overlapping bounding boxes. In this paper, we address the weaknesses of $IoU$ by introducing a generalized version as both a new loss and a new metric. By incorporating this generalized $IoU$ ($GIoU$) as a loss into the state-of-the art object detection frameworks, we show a consistent improvement on their performance using both the standard, $IoU$ based, and new, $GIoU$ based, performance measures on popular object detection benchmarks such as PASCAL VOC and MS COCO.

In object detection algorithms, the bounding box regression (BBR) loss directly influences the model's accuracy in predicting object positions. Initially, we conducted simulation experiments on the proposed intersection over union-based loss. Through in-depth analysis, we discovered that the width and height components of the distance term could hinder BBR. To address this, we proposed the AIoU-v1 loss, which decouples the width and height components of the distance term, thereby preventing the suppression of BBR loss by the distance term. Additionally, issues such as the imbalance between sample quantity and sample quality in the dataset, as well as labeling errors, can adversely affect BBR. To tackle these dataset problems, we designed an adaptive dynamic non-monotonic focusing mechanism with strong robustness and wide applicability. Finally, we proposed a post-processing algorithm that combines fusion and non-maximum suppression, resulting in more accurate bounding boxes during the post-processing stage. Our source code and data are available at https://github.com/Wenhua-Zheng/AIOU.

Object detection and recognition is an essential task in computer vision with numerous real-world applications such as surveillance, self-driving cars, and robotics. In recent years, deep learning-based approaches have significantly improved the accuracy and speed of object detection and recognition. The You Only Look Once version 3 (YOLOv3) algorithm is a popular deep learning-based approach that can detect and recognize objects in real-time. The Common Objects in Context (COCO) dataset is a large-scale dataset with over 330,000 labeled images and more than 2.5 million object instances, making it a popular choice for object detection and recognition tasks. In this paper, we propose a deep learning-based approach for real-time object detection and recognition using the YOLOv3 architecture and COCO dataset. We evaluate our approach based on several performance metrics, including mean average precision (mAP), frames per second (FPS), total object detection time, object detection accuracy, false positive rate, number of detected objects, and mean intersection over union (mIoU). Our results show that our approach achieves a mean average precision of 0.76 on the COCO dataset and a real-time performance of 40 frames per second on a single GPU. Additionally, our approach achieves an object detection accuracy of 93.5%, a false positive rate of 6.5%, and a mean intersection over union of 0.65. Our proposed approach shows promising results for real-time object detection and recognition and can be applied to various real-world applications.

Object detection is a cornerstone task in computer vision, impacting diverse domains from autonomous systems to medical imaging. This study presents a systematic evaluation of three leading detection architecturesFaster Region-based Convolutional Neural Networks (Faster R-CNN), You Only Look Once version 8 (YOLOv8), and Detection Transformer (DETR), with the objective of evaluating their performance characteristics and practical deployment potential. Through comparative experiments, the research evaluates Faster R-CNNs precision (Delivers 76.4% mean Average Precision with a processing speed of 5-7 Frames Per Second (FPS)), YOLOv8s real-time efficiency (53.9% mean Average Precision (mAP) at 80+ FPS), and DETR attention-based innovation (42% AP). Results highlight the strengths and weaknesses of each model: Faster R-CNN excels in accuracy-demanding applications like medical diagnosis, YOLOv8 dominates real-time tasks such as autonomous driving, and DETR offers promising temporal analysis capabilities despite higher computational costs. The study proposes innovative solutions, including cross-scale attention mechanisms and dynamic inference techniques, to address limitations such as small-object detection and edge deployment. The findings provide valuable insights for architecture selection, offering actionable guidelines for industrial implementation and laying the groundwork for future advancements in multimodal fusion and self-supervised learning paradigms. This framework accelerates the development of next-generation detection systems tailored for emerging Artificial Intelligence (AI) applications.

Object detection is a fundamental task in computer vision, which is usually based on convolutional neural networks (CNNs). While it is difficult to be deployed in embedded devices due to the huge storage and computing consumptions, binary neural networks (BNNs) can execute object detection with limited resources. However, the extreme quantification in BNN causes diversity of feature representation loss, which eventually influences the object detection performance. In this paper, we propose a method balancing Information Retention and Deviation Control to achieve effective object detection, named IR-DC Net. On the one hand, we introduce the KL-Divergence to compose multiple entropy for maximizing the available information. On the other hand, we design a lightweight convolutional module to generate scale factors dynamically for minimizing the deviation between binary and real convolution. The experiments on PASCAL VOC, COCO2014, KITTI, and VisDrone datasets show that our method improved the accuracy in comparison with previous binary neural networks.

Object detection is one of the essential tasks in computer vision, with most detection methods relying on a limited number of sizes for anchor boxes. However, the boundaries of particular composite objects, such as ports, highways, and golf courses, are ambiguous in remote sensing images, and therefore, it is challenging for the anchor-based method to accommodate the substantial size variation of the objects. In addition, the dense placement of anchor boxes imbalances the positive and negative samples, which affects the end-to-end architecture of deep learning methods. Hence, this paper proposes a single-stage object detection model named Xnet to address this issue. The proposed method designs a deformable convolution backbone network used in the feature extraction stage. Compared to the standard convolution, it adds learnable parameters for dynamically analyzing the boundary and offset of the receptive field, rendering the model more adaptable to size variations within the same class. Moreover, this paper presents a novel anchor-free detector that classifies objects in feature images point-by-point, without relying on anchor boxes. Several experiments on the large remote sensing dataset DIOR challenging Xnet against other popular methods demonstrate that our method attains the best performance, surpassing by 4.7% on the mAP (mean average precision) metric.

Recently, most state-of-the-art object detection systems adopt anchor box mechanism to simplify the detection model. Neural networks only need to regress the mapping relations from anchor boxes to ground truth boxes, then prediction boxes can be calculated using information from outputs of networks and default anchor boxes. However, when the problem becomes complex, the number of default anchor boxes will increase with large risk of over-fitting during training. In this paper, we adopt an adaptive anchor box mechanism that one anchor box can cover more ground truth boxes. So networks only need a few adoptive anchor boxes to solve the same problem and the model will be more robust. The sizes of adaptive anchor boxes will be adjusted automatically according to the depth collected by a Time of Flight (TOF) camera. The network adjusts the aspect ratios of anchor boxes to get final prediction boxes. The experimental results demonstrate that the proposed method can get more accurate detection results. Specifically, using the proposed adaptive anchor box mechanism, the Mean Average Precision (mAP) of YOLO-v2 and YOLO-v3 networks increases obviously on open public datasets and our self-built battery image dataset. Moreover, the visual results of prediction comparisons also illustrate that the proposed adaptive anchor box mechanism can achieve better performance than original anchor box mechanism.

Object detection in aerial images is an active yet challenging task in computer vision because of the bird’s-eye view perspective, the highly complex backgrounds, and the variant appearances of objects. Especially when detecting densely packed objects in aerial images, methods relying on horizontal proposals for common object detection often introduce mismatches between the Region of Interests (RoIs) and objects. This leads to the common misalignment between the final object classification confidence and localization accuracy. In this paper, we propose a RoI Transformer to address these problems. The core idea of RoI Transformer is to apply spatial transformations on RoIs and learn the transformation parameters under the supervision of oriented bounding box (OBB) annotations. RoI Transformer is with lightweight and can be easily embedded into detectors for oriented object detection. Simply apply the RoI Transformer to light head RCNN has achieved state-of-the-art performances on two common and challenging aerial datasets, i.e., DOTA and HRSC2016, with a neglectable reduction to detection speed. Our RoI Transformer exceeds the deformable Position Sensitive RoI pooling when oriented bounding-box annotations are available. Extensive experiments have also validated the flexibility and effectiveness of our RoI Transformer.

Siamese network-based visual tracking algorithms have achieved excellent performance in recent years, but challenges such as fast target motion, shape and scale variations have made the tracking extremely difficult. The regression of anchor-free tracking has low computational complexity, strong real-time performance, and is suitable for visual tracking. Based on the anchor-free siamese tracking framework, this paper firstly introduces balance factors and modulation coefficients into the cross-entropy loss function to solve the classification inaccuracy caused by the imbalance between positive and negative samples as well as the imbalance between hard and easy samples during the training process, so that the model focuses more on the positive samples and the hard samples that make the major contribution to the training. Secondly, the intersection over union (IoU) loss function of the regression branch is improved, not only focusing on the IoU between the predicted box and the ground truth box, but also considering the aspect ratios of the two boxes and the minimum bounding box area that accommodate the two, which guides the generation of more accurate regression offsets. The overall loss of classification and regression is iteratively minimized and improves the accuracy and robustness of visual tracking. Experiments on four public datasets, OTB2015, VOT2016, UAV123 and GOT-10k, show that the proposed algorithm achieves the state-of-the-art performance.

A multi-modal 3D object-detection method, based on data from cameras and LiDAR, has become a subject of research interest. PointPainting proposes a method for improving point-cloud-based 3D object detectors using semantic information from RGB images. However, this method still needs to improve on the following two complications: first, there are faulty parts in the image semantic segmentation results, leading to false detections. Second, the commonly used anchor assigner only considers the intersection over union (IoU) between the anchors and ground truth boxes, meaning that some anchors contain few target LiDAR points assigned as positive anchors. In this paper, three improvements are suggested to address these complications. Specifically, a novel weighting strategy is proposed for each anchor in the classification loss. This enables the detector to pay more attention to anchors containing inaccurate semantic information. Then, SegIoU, which incorporates semantic information, instead of IoU, is proposed for the anchor assignment. SegIoU measures the similarity of the semantic information between each anchor and ground truth box, avoiding the defective anchor assignments mentioned above. In addition, a dual-attention module is introduced to enhance the voxelized point cloud. The experiments demonstrate that the proposed modules obtained significant improvements in various methods, consisting of single-stage PointPillars, two-stage SECOND-IoU, anchor-base SECOND, and an anchor-free CenterPoint on the KITTI dataset.

Object detection is a challenging task in the field of remote sensing applications due to the complex backgrounds and uncertain orientation of targets. Compared with the horizontal bounding box, the oriented bounding box can provide orientation information while retaining the true size. Most existing oriented object detection methods are based on Faster-RCNN and the other one-stage methods that can achieve real-time speed but have shortcomings in localization and detection accuracy. To further enhance the performance of one-stage methods, we propose an oriented object detection framework that is based on the single shot detector, namely, single shot anchor refinement network (S²ARN). The S²ARN obtains the accurate detection results by performing two consecutive regressions. More precisely, the multilevel features of the backbone are used to regress the coordinate offsets between the predefined rotated anchors and the ground-truth boxes to generate the refined anchors. The classification and regression subnetworks assigned to the output features are used to perform the second regression to determine the class labels and further adjust the location of the refined anchors. In addition, receptive field amplification modules (RFAMs) are inserted to enlarge the receptive field and extract more discriminative features. Furthermore, in the anchor matching step, angle-related Intersection over Union (ArIoU) is used to calculate the Intersection over Union (IoU) score instead of the traditional method. Benefiting from the multiple regressions and the insensitivity of the ArIoU score to the angle deviation, the angle sampling interval of the rotated anchor can be reduced. The experimental results for the two public datasets, HRSC2016 and UCAS-AOD, demonstrate the effectiveness of the proposed network.

A challenging and attractive task in computer vision is underwater object detection. Although object detection techniques have achieved good performance in general datasets, problems of low visibility and color bias in the complex underwater environment have led to generally poor image quality; besides this, problems with small targets and target aggregation have led to less extractable information, which makes it difficult to achieve satisfactory results. In past research of underwater object detection based on deep learning, most studies have mainly focused on improving detection accuracy by using large networks; the problem of marine underwater lightweight object detection has rarely gotten attention, which has resulted in a large model size and slow detection speed; as such the application of object detection technologies under marine environments needs better real-time and lightweight performance. In view of this, a lightweight underwater object detection method based on the MobileNet v2, You Only Look Once (YOLO) v4 algorithm and attentional feature fusion has been proposed to address this problem, to produce a harmonious balance between accuracy and speediness for target detection in marine environments. In our work, a combination of MobileNet v2 and depth-wise separable convolution is proposed to reduce the number of model parameters and the size of the model. The Modified Attentional Feature Fusion (AFFM) module aims to better fuse semantic and scale-inconsistent features and to improve accuracy. Experiments indicate that the proposed method obtained a mean average precision (mAP) of 81.67% and 92.65% on the PASCAL VOC dataset and the brackish dataset, respectively, and reached a processing speed of 44.22 frame per second (FPS) on the brackish dataset. Moreover, the number of model parameters and the model size were compressed to 16.76% and 19.53% of YOLO v4, respectively, which achieved a good tradeoff between time and accuracy for underwater object detection.

Multi-label image classification is a fundamental but challenging task in computer vision. To tackle the problem, the label-related semantic information is often exploited, but the background context and spatial semantic information of related objects are not fully utilized. To address these issues, a multi-branch deep neural network is proposed in this paper. The first branch is designed to extract the discriminant information from regions of interest to detect target objects. In the second branch, a spatial context-aware approach is proposed to better capture the contextual information of an object in its surroundings by using an adaptive patch expansion mechanism. It helps the detection of small objects that are easily lost without the support of context information. The third one, the object-attentional branch, exploits the spatial semantic relations between the target object and its related objects, to better detect partially occluded, small or dim objects with the support of those easily detectable objects. To better encode such relations, an attention mechanism jointly considering the spatial and semantic relations between objects is developed. Two widely used benchmark datasets for multi-labeling classification, MS COCO and PASCAL VOC, are used to evaluate the proposed framework. The experimental results demonstrate that the proposed method outperforms the state-of-the-art methods for multi-label image classification.

Object detection from UAV thermal infrared images and videos using YOLO models