Dynamic Anchor Learning for Arbitrary-Oriented Object Detection

Gefan

IEEE Trans. Pattern Anal. Mach. Intell.

et al. 2022

Existing detection methods commonly use a parameterized bounding box (BBox) to model and detect (horizontal) objects and an additional rotation angle parameter is used for rotated objects. We argue that such a mechanism has fundamental limitations in building an effective regression loss for rotation detection, especially for high-precision detection with high IoU (e.g. 0.75). Instead, we propose to model the rotated objects as Gaussian distributions. A direct advantage is that our new regression loss regarding the distance between two Gaussians e.g. Kullback-Leibler Divergence (KLD), can well align the actual detection performance metric, which is not well addressed in existing methods. Moreover, the two bottlenecks i.e. boundary discontinuity and square-like problem also disappear. We also propose an efficient Gaussian metric-based label assignment strategy to further boost the performance. Interestingly, by analyzing the BBox parameters' gradients under our Gaussian-based KLD loss, we show that these parameters are dynamically updated with interpretable physical meaning, which help explain the effectiveness of our approach, especially for high-precision detection. We extend our approach from 2-D to 3-D with a tailored algorithm design to handle the heading estimation, and experimental results on twelve public datasets (2-D/3-D, aerial/text/face images) with various base detectors show its superiority.

“…Extending the above horizontal case, existing rotation detection models [1], [12], [17], [36], [57] also use regression loss which simply involves an extra angle parameter θ:…”

Section: Regression Loss Design Revisit: From Horizon To Rotation Det...mentioning

confidence: 99%

Section: Kullback-leibler Divergencementioning

confidence: 99%

Detecting Rotated Objects as Gaussian Distributions and Its 3-D Generalization

Gefan

IEEE Trans. Pattern Anal. Mach. Intell.

et al. 2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

“…CAD-Net [10] learns global and local contextual information of objects by computing their correlations with the global scene and local adjacent features. DAL [13] is a dynamic anchor learning method that uses a new matching mechanism to evaluate anchors and assign them more efficient labels. S 2 A-Net [14] uses a new alignment convolution, which can adaptively align convolution features according to anchors.…”

Section: Comparisons With the State-of-the-artmentioning

confidence: 99%

“…1(a), HBBs representing rotated objects may contain a lot of undesirable contents such as a large amount of background for narrow objects with large aspect ratios and parts of other objects for densely distributed objects. For better localization of rotated objects, oriented bounding boxes (OBBs) are widely used in RSIs objects detection [9][10][11][12][13][14][15]. As can be seen in Fig.…”

mentioning

confidence: 99%

Orientation-First Strategy With Angle Attention Module for Rotated Object Detection in Remote Sensing Images

Zhang

Wang

Zhang

et al. 2022

Recently, object detection in remote sensing images (RSIs) have received extensive attention and made significant progress. Nonetheless, the arbitrary orientations of objects in RSIs make their detection a challenging task. Most of the existing detection methods are difficult to extract the orientation features of objects due to the lack of directionality of conventional convolutions. In addition, the boundary discontinuity in angle regression affects the detection of object orientations. In response to these problems, this paper proposes an orientation-first refinement detector (OFRDet), which is based on a strategy that enables the detector to detect the angle of an object ahead of others and presets oriented anchors. In OFRDet, we propose an angle encoding regression module (AERM) and an angle channel attention module (ACAM). AERM transforms angle detection into multi-parameter regression, which eliminates boundary discontinuities. ACAM uses convolution kernels with different angles to extract directional features purposefully according to the preset oriented anchors. After these two modules, more accurate bounding boxes are generated and sent to the refined stage to obtain the final detection results. We evaluate our method and demonstrate the effectiveness of it by conducting experiments on two challenging and credible datasets, DOTA, HRSC2016. OFRDet achieves competitive results 79.56%, 96.29% mAP on the two datasets, respectively. Index Terms-Angle channel attention, angle encoding, remote sensing images, rotated object detection I. INTRODUCTION BJECT detection is a technique in computer vision that requires locating and identifying the certain object in the image. Remote sensing images (RSIs) are more challenging to be detected since the scale of RSIs is larger and the content is more complex than that of ordinary natural images [1]. In addition, objects are unevenly distributed on RSIs and are generally small. With the continuous development of deep learning technology, neural networks are widely used in image processing. Meanwhile the object detection based on convolutional neural networks (CNNs) have made great Manuscript

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

“…After selecting a fixed number of positive candidates, the cleanliness score is used to generate soft labels for classification and weights of loss terms for regression. Similarly, Dynamic Anchor Learning (DAL) [54] selects the candidates according to the prediction results. In addition to output IoU, it also considers the input IoU to stabilize the training.…”

Section: B Label Assignmentmentioning

confidence: 99%

Constrained-SIoU: A Metric for Horizontal Candidates in Multi-Oriented Object Detection

Zhang

Wang

et al. 2022

Intersection over union (IoU) has been widely adopted to evaluate and select candidate regions in multi-oriented object detection. Intuitively, overlaps between candidates and multioriented ground-truth boxes make more sense when assessing the quality of horizontal candidates. However, the horizontal minimum bounding box (HMBB) of the ground-truth box is generally used for the IoU calculation in practice, bringing about biased results. In this article, we propose a novel Splicing Intersection over Union (SIoU) to provide a more preferable metric for horizontal candidate selection when detecting multi-oriented objects. By computing the intersection between the candidate region and the ground-truth box rather than its HMBB, SIoU provides a better description of how much object information a candidate contains. Furthermore, we introduce two variants of constraints for the center of each candidate to ensure its location focusing on the objects. Candidates whose centers deviate too far from the objects will be penalized. We integrate the constraint with SIoU, denoted as constrained-SIoU, to select horizontal candidates more efficiently. Comparative experiments on two datasets of aerial images, DOTA and HRSC2016, demonstrate the effectiveness of the proposed method.