Dynamic Anchor Learning for Arbitrary-Oriented Object Detection

By Qi Ming, Zhiqiang Zhou, Lingjuan Miao, Hongwei Zhang, Linhao Li. AAAI Conference on Artificial Intelligence, 2020

343 citations recorded by the source corpus.

Arbitrary-oriented objects widely appear in natural scenes, aerial photographs, remote sensing images, etc., and thus arbitrary-oriented object detection has received considerable attention. Many current rotation detectors use plenty of anchors with different orientations to achieve spatial alignment with ground truth boxes. Intersection-over-Union (IoU) is then applied to sample the positive and negative candidates for training. However, we observe that the selected positive anchors cannot always ensure accurate detections after regression, while some negative samples can achieve accurate localization. It indicates that the quality assessment of anchors through IoU is not appropriate, and this further leads to inconsistency between classification confidence and localization accuracy. In this paper, we propose a dynamic anchor learning (DAL) method, which utilizes the newly defined matching degree to comprehensively evaluate the localization potential of the anchors and carries out a more efficient label assignment process. In this way, the detector can dynamically select high-quality anchors to achieve accurate object detection, and the divergence between classification and regression will be alleviated. With the newly introduced DAL, we can achieve superior detection performance for arbitrary-oriented objects with only a few horizontal preset anchors. Experimental results on three remote sensing datasets HRSC2016, DOTA, UCAS-AOD as well as a scene text dataset ICDAR 2015 show that our method achieves substantial improvement compared with the baseline model. Besides, our approach is also universal for object detection using horizontal bound box. The code and models are available at https://github.com/ming71/DAL.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/5cc2198ddbbec84f04019b1fd3f02798673dbb91

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Junwei Han
2. Gong Cheng
3. Jianke Zhu
4. Wentong Li
5. Xiwen Yao
6. Junchi Yan
7. Xue Yang
8. Ke Li

In Media Technology

1. Gong Cheng
2. Junwei Han
3. Xiwen Yao
4. Ke Li
5. Jianke Zhu
6. Wentong Li
7. Xingxing Xie
8. Xue Yang

Citations in the corpus

• Anchor-Free Oriented Proposal Generator for Object Detection (2021, background): “RRPN [6] R2CNN [41] RT [1] GV [2] DRN [21] PIoU [38] DAL [39] S(2)ANet [4] AOPG”
• Poly Kernel Inception Network for Remote Sensing Detection (2024, methodology): “This includes introducing feature refinement techniques into the detector neck [69, 71], extracting the rotated region of interest (RoI) [10, 65], designing specific detection heads for the OBBs [21, 26, 48], .etc .”
• MMRotate: A Rotated Object Detection Benchmark using PyTorch (2022, background): “DAL [22] reconsiders whether IoU is a truly”
• Oriented R-CNN for Object Detection (2021, methodology): “In addition, some works [12, 28, 39, 27, 43, 16, 40, 47, 15] have explored one-stage or anchor-free oriented object detection frameworks: outputting object classes and oriented bounding boxes without region proposal generation and RoI alignment operation.”
• Oriented RepPoints for Aerial Object Detection (2021, background): “[23] introduce a matching degree measure to evaluate the spatial alignment based on the oriented anchors, which use the matching sensitive loss to enhance the correlation between classification and oriented localization.”