Abstract:In aerial image scenes, the objects have properties of arbitrary orientation, large-scale range, and dense distribution. Thus, the object detector uses an oriented bounding box (OBB) to locate objects, which is more complex and challenging than a horizontal bounding box (HBB) detector. Mainstream OBB detectors mostly use a one-to-many label assignment strategy to predict multiple bounding boxes for the same object and filter out repeat predictions by nonmaximum suppression (NMS). NMS ranks with confidence and drops the detection box with intersection over union (IoU) higher than the threshold, which makes it easy to get the local optimum result. The clustered synthesis method gets more accurate results than the original NMS, but applying it to the OBB detector leads to border shift, which arises from the angular discontinuity problem. Therefore, we use Gaussian OBB (G-OBB) to deal with the angular discontinuity and thus eliminate the offset generated by direct synthesis. G-OBB is not easy to understand and describe representation. For this reason, we analyze the properties of G-OBB and design a decoding method to convert a G-OBB to a rotated rectangular box, further discussing its conditions. Based on the decoding method, we propose a Gaussian synthesis (GauS) algorithm, which transforms the OBB into Gaussian space, followed by synthesis, and finally transforms the synthesis result back into a new OBB. We have derived the synthesis and decoding methods and further verified their effectiveness. The extensive experiments on several existing models show that GauS takes very little computation and improves the detector’s high-precision performance. Extensive experiments verify the effectiveness, stability, and universality of the proposed algorithm. In addition, the RTMDet using GauS achieves a performance of 81.61 AP50 and gains a 0.39% improvement in mean average precision (mAP), which achieves the state-of-the-art (SOTA) performance. Our implementation is available at https://github.com/lzh420202/GauS .

Visual oriented object detection via feature alignment and Gaussian parameterization

Oriented Object Detector with Gaussian Distribution Cost Label Assignment and Task-Decoupled Head

A Refined Single-Stage Detector With Feature Enhancement and Alignment for Oriented Objects

Dual-Aligned Oriented Detector

An Arbitrary-Oriented Object Detector Based on Variant Gaussian Label in Remote Sensing Images

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

Gaussian Aware Anchor-Free Rotated Detector for Aerial Object Detection

Oriented Object Detection via Contextual Dependence Mining and Penalty-Incentive Allocation

Recalibrating Features and Regression for Oriented Object Detection.

On Improving Bounding Box Representations for Oriented Object Detection

MRDet: A Multi-Head Network for Accurate Oriented Object Detection in Aerial Images

Feature Enhancement and Alignment for Oriented Object Detection

Precise Vertex Regression and Feature Decoupling for Oriented Object Detection

M2VDet: Midpoints-to-Vertices Detection of Oriented Objects in Remote-Sensing Images.

Align Deep Features for Oriented Object Detection

Shared-Weight-Based Multi-Dimensional Feature Alignment Network for Oriented Object Detection in Remote Sensing Imagery

Aspect-Ratio-Guided Detection for Oriented Objects in Remote Sensing Images

Gaussian Synthesis for High-Precision Location in Oriented Object Detection.

Highly Efficient Anchor-Free Oriented Small Object Detection for Remote Sensing Images via Periodic Pseudo-Domain

Structurally re-parameterized rotation detector for arbitrary-oriented objects in high-resolution remote sensing images

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