Dynamic Anchor Learning for Arbitrary-Oriented Object Detection

Qi, M.; Zhou, Zhiqiang; Miao, Lingjuan; Zhang, Hongwei; Li, Linhao

doi:10.48550/arxiv.2012.04150

Cited by 11 publications

(19 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Benefiting from the open-source AOOD datasets annotated with OBBs in the scenes like remote sensing [5], the prediction of the OD model has become more refined, which helps to accurately locate the object in the image and reflect its shape and direction. In the AOOD task, whether the two-stage methods [10,20,21] or the one-stage methods [7,11,12], most of them adopt the anchor-box-based framework due to its mature application in various OD tasks. However, since oriented anchors are more prone to mismatch problems and have more hyperparameters than horizontal anchors, many works have dealt with them.…”

Section: A Arbitrary-oriented Object Detectionmentioning

confidence: 99%

“…Xu et al [21] proposed a gliding vertex method to represent OBBs, the model of which is based on horizontal anchors without setting oriented anchors with multiple angles. DAL [12] analyzed and proposed a dynamic matching and assignment strategy. To remove anchor boxes, BBAVectors [17], DRN [7], O 2 -DNet [18], etc., employed the anchor-free framework and designed new OBB representation components.…”

Section: A Arbitrary-oriented Object Detectionmentioning

confidence: 99%

“…Most of the AOOD methods, such as SCRDet [10], LO-Det [11], DAL [12], etc., use the anchor-based label assignment strategy, as shown in Fig. 2 (a).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A General Gaussian Heatmap Label Assignment for Arbitrary-Oriented Object Detection

Huang,

Li,

Xia

et al. 2021

Preprint

View full text Add to dashboard Cite

Recently, many arbitrary-oriented object detection (AOOD) methods have been proposed and attracted widespread attention in many fields. However, most of them are based on anchor-boxes or standard Gaussian heatmaps. Such label assignment strategy may not only fail to reflect the shape and direction characteristics of arbitrary-oriented objects, but also have high parameter-tuning efforts. In this paper, a novel AOOD method called General Gaussian Heatmap Labeling (GGHL) is proposed. Specifically, an anchor-free object-adaptation label assignment (OLA) strategy is presented to define the positive candidates based on two-dimensional (2-D) oriented Gaussian heatmaps, which reflect the shape and direction features of arbitrary-oriented objects. Based on OLA, an oriented-boundingbox (OBB) representation component (ORC) is developed to indicate OBBs and adjust the Gaussian center prior weights to fit the characteristics of different objects adaptively through neural network learning. Moreover, a joint-optimization loss (JOL) with area normalization and dynamic confidence weighting is designed to refine the misalign optimal results of different subtasks. Extensive experiments on public datasets demonstrate that the proposed GGHL improves the AOOD performance with low parameter-tuning and time costs. Furthermore, it is generally applicable to most AOOD methods to improve their performance including lightweight models on embedded platforms.

show abstract

Section: A Arbitrary-oriented Object Detectionmentioning

confidence: 99%

Section: A Arbitrary-oriented Object Detectionmentioning

confidence: 99%

See 1 more Smart Citation

A General Gaussian Heatmap Label Assignment for Arbitrary-Oriented Object Detection

Huang,

Li,

Xia

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Arbitrary-oriented object detection has a wide range of application scenarios, such as scene text detection [18,11,6], face detection [27], object detection in remote sensing images [34,8,35,30,39,2,38], and 3D object detection [45]. In recent years, with the breakthroughs made by convolutional neural networks (CNNs) in the field of object detection [4,26,15,24], a series of CNN-based rotation detectors have been proposed to achieve high-performance oriented object detection [36,20,19,16,31]. 𝑔 0 : (0, 0, 100, 600, −30 𝑜 ) 𝑔 1 : (0, 0, 600, 100, −120 𝑜 ) 𝑎: (0, 0, 600, 100, −90 𝑜 ) Unlike generic object detection that uses the horizontal bounding box to represent the objects, rotation detectors usually adopt the oriented bounding box (OBB) [2,20,19] or the quadrilateral bounding box (QBB) [31,11,16] to describe the rotating objects, which induces the representation ambiguity.…”

Section: Introductionmentioning

confidence: 99%

Optimization for Arbitrary-Oriented Object Detection via Representation Invariance Loss

Zhou

Miao

et al. 2022

IEEE Geosci. Remote Sensing Lett.

Self Cite

View full text Add to dashboard Cite

Arbitrary-oriented objects exist widely in natural scenes, and thus the oriented object detection has received extensive attention in recent years. The mainstream rotation detectors use oriented bounding boxes (OBB) or quadrilateral bounding boxes (QBB) to represent the rotating objects. However, these methods suffer from the representation ambiguity for oriented object definition, which leads to suboptimal regression optimization and the inconsistency between the loss metric and the localization accuracy of the predictions. In this paper, we propose a Representation Invariance Loss (RIL) to optimize the bounding box regression for the rotating objects. Specifically, RIL treats multiple representations of an oriented object as multiple equivalent local minima, and hence transforms bounding box regression into an adaptive matching process with these local minima. Then, the Hungarian matching algorithm is adopted to obtain the optimal regression strategy. We also propose a normalized rotation loss to alleviate the weak correlation between different variables and their unbalanced loss contribution in OBB representation. Extensive experiments on remote sensing datasets and scene text datasets show that our method achieves consistent and substantial improvement. The source code and trained models are available at https://github.com/ming71/RIDet.

show abstract

“…), and utilizes a rotation box to locate the object (see Figure 1). Many researchers have proposed different rotating object detection methods based on R-CNN frameworks [5,39,33,36] or one-stage method frameworks [34,7,24]. Compared with horizontal detection, rotation detection seems more difficult to select positive samples from horizontal anchors, which should have large Intersection-over-Union (IoU) with the ground-truth.…”

Section: Introductionmentioning

confidence: 99%

TS4Net: Two-Stage Sample Selective Strategy for Rotating Object Detection

Feng¹,

Li²,

Han³

et al. 2021

Preprint

View full text Add to dashboard Cite

Rotating object detection has wide applications in aerial photographs, remote sensing images, UAVs, etc. At present, most of the rotating object detection datasets focus on the field of remote sensing, and these images are usually shot in high-altitude scenes. However, image datasets captured at low-altitude areas also should be concerned, such as drone-based datasets. So we present a low-altitude dronebased dataset, named UAV-ROD, aiming to promote the research and development in rotating object detection and UAV applications. The UAV-ROD consists of 1577 images and 30,090 instances of car category annotated by oriented bounding boxes. In particular, The UAV-ROD can be utilized for the rotating object detection, vehicle orientation recognition and object counting tasks. Compared with horizontal object detection, the regression stage of the rotation detection is a tricky problem. In this paper, we propose a rotating object detector TS 4 Net, which contains anchor refinement module (ARM) and two-stage sample selective strategy (TS 4 ). The ARM can convert preseted horizontal anchors into high-quality rotated anchors through twostage anchor refinement. The TS 4 module utilizes different constrained sample selective strategies to allocate positive and negative samples, which is adaptive to the regression task in different stages. Benefiting from the ARM and TS 4 , the TS 4 Net can achieve superior performance for rotating object detection solely with one preseted horizontal anchor. Extensive experimental results on UAV-ROD dataset and three remote sensing datasets DOTA, HRSC2016 and UCAS-AOD demonstrate that our method achieves competitive performance against most state-of-the-art methods.

show abstract

Dynamic Anchor Learning for Arbitrary-Oriented Object Detection

Cited by 11 publications

References 42 publications

A General Gaussian Heatmap Label Assignment for Arbitrary-Oriented Object Detection

A General Gaussian Heatmap Label Assignment for Arbitrary-Oriented Object Detection

Optimization for Arbitrary-Oriented Object Detection via Representation Invariance Loss

TS4Net: Two-Stage Sample Selective Strategy for Rotating Object Detection

Contact Info

Product

Resources

About