Arbitrary-oriented object detection is a fundamental task in visual scenes involving aerial images and scene text. In this report, we present PP-YOLOE-R, an efficient anchor-free rotated object detector based on PP-YOLOE. We introduce a bag of useful tricks in PP-YOLOE-R to improve detection precision with marginal extra parameters and computational cost. As a result, PP-YOLOE-R-l and PP-YOLOE-R-x achieve 78.14 and 78.28 mAP respectively on DOTA 1.0 dataset with single-scale training and testing, which outperform almost all other rotated object detectors. With multi-scale training and testing, PP-YOLOE-R-l and PP-YOLOE-R-x further improve the detection precision to 80.02 and 80.73 mAP. In this case, PP-YOLOE-R-x surpasses all anchor-free methods and demonstrates competitive performance to state-of-the-art anchor-based two-stage models. Further, PP-YOLOE-R is deployment friendly and PP-YOLOE-R-s/m/l/x can reach 69.8/55.1/48.3/37.1 FPS respectively on RTX 2080 Ti with TensorRT and FP16-precision. Source code and pre-trained models are available at https://github.com/PaddlePaddle/PaddleDetection, which is powered by https://github.com/PaddlePaddle/Paddle.
翻译:在空中图像和场景文字的视觉场景中,以任意为导向的物体探测是一项基本任务。在本报告中,我们介绍了PP-YOLOE-R,这是一个基于PP-YOLOE的有效无锚旋转物体探测器。我们在PP-YOLOOE-R中引入了一袋有用的技巧,用边际额外参数和计算成本来提高探测精确度。因此,PP-YOLOOE-R-l和PP-YOLOE-R-x在DOTA 1.0数据集上分别实现了78.14和78.28 mAP,采用单级培训和测试,几乎超越了所有其他旋转物体探测器。在多级培训和测试中,PPP-YOLOE-R-l和PP-YOLOE-R-x中引入了一袋有用的技巧,使探测精确精确度达到8002和80.73 mAPL.P-PA-P-PAL-PLE-PAL-S-S-PLS-PLS-PLS-PLES-PLA/PLE-PLE/PLAFS-PS-PLA/P-PLES-PLE/PLE/PLS-PS-PLS-PLS-PLS-S-S-S-PLS-PLS-S-S-S-S-S-PS-S-S-S-S-PLS-PLS-PLS-S-S-S-S-S-PLS-PLS-PLS-PLS-PLS-S-S-S-PLS-S-S-S-S-S-S-S-S-S-S-S-S-S-S-S-S-S-S-S-P-S-P-S-S-PLS-S-PLS-PLS-S-S-PLS-S-S-PLS-PLS-S-S-S-S-S-PLS-S-PLS-PLS-PLS-S-S-PLS-S-S-S-S-S-P