ReDet: A Rotation-equivariant Detector for Aerial Object Detection

By Jiaming Han, Jian Ding, Nan Xue, Guisong Xia. Computer Vision and Pattern Recognition, 2021

830 citations recorded by the source corpus.

Recently, object detection in aerial images has gained much attention in computer vision. Different from objects in natural images, aerial objects are often distributed with arbitrary orientation. Therefore, the detector requires more parameters to encode the orientation information, which are often highly redundant and inefficient. Moreover, as ordinary CNNs do not explicitly model the orientation variation, large amounts of rotation augmented data is needed to train an accurate object detector. In this paper, we propose a Rotation-equivariant Detector (ReDet) to address these issues, which explicitly encodes rotation equivariance and rotation invariance. More precisely, we incorporate rotation-equivariant networks into the detector to extract rotation-equivariant features, which can accurately predict the orientation and lead to a huge reduction of model size. Based on the rotation-equivariant features, we also present Rotation-invariant RoI Align (RiRoI Align), which adaptively extracts rotation-invariant features from equivariant features according to the orientation of RoI. Extensive experiments on several challenging aerial image datasets DOTA-v1.0, DOTA-v1.5 and HRSC2016, show that our method can achieve state-of-the-art performance on the task of aerial object detection. Compared with previous best results, our ReDet gains 1.2, 3.5 and 2.6 mAP on DOTA-v1.0, DOTA-v1.5 and HRSC2016 respectively while reducing the number of parameters by 60% (313 Mb vs. 121 Mb). The code is available at: https://github.com/csuhan/ReDet.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/0fdb3478b2db17de56ff9c7d338e95559a4caf12

How researchers see this paper

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

In Aerospace Engineering

1. Wentong Li
2. Qinghua Hu
3. Pengfei Zhu
4. Jianke Zhu
5. Yiming Sun
6. Bing Cao

In Computer Vision and Pattern Recognition

1. Wentong Li
2. Qinghua Hu
3. Pengfei Zhu
4. Jianke Zhu
5. Yiming Sun
6. Xinhao Cai
7. Yazhou Yao
8. Qiuxia Lai

Citations in the corpus

• Poly Kernel Inception Network for Remote Sensing Detection (2024, methodology): “Some [21, 50] focus on extracting features suitable for objects of varying orientations with equivalent receptive fields.”
• MMRotate: A Rotated Object Detection Benchmark using PyTorch (2022, neutral): “…range, it can be divided into OpenCV definition ( 𝐷 𝑜𝑐 ,𝜃 ∈ [− 𝜋 / 2 , 0 ) ) [32, 36, 38], long edge 90° definition ( 𝐷 𝑙𝑒 90 ,𝜃 ∈ [− 𝜋 / 2 , 𝜋 / 2 ) ) [3, 6], and long edge 135° ) ) [5]. https://github.com/dingjiansw101/AerialDetection 4 https://github.com/jbwang1997/OBBDetection Recent…”
• Drone-Based RGB-Infrared Cross-Modality Vehicle Detection Via Uncertainty-Aware Learning (2020, methodology): “637 2) Baselines: For the DroneVehicle dataset, 638 we compared our method with nine competing methods: 639 RetinaNet(OBB) [42], Faster R-CNN(OBB) [8], Faster 640 R-CNN(Dpool) [53], Mask R-CNN [45], Cascade Mask 641 R-CNN [46], Hybrid Task Cascade [47], RoITransformer [9], 642 ReDet [54], and Glidin”
• Oriented RepPoints for Aerial Object Detection (2021, methodology): “The mainstreamed approaches typically treat aerial object detection as a problem of rotated object localization [3, 8,9,24,42,45].”