Mask OBB: A Semantic Attention-Based Mask Oriented Bounding Box Representation for Multi-Category Object Detection in Aerial Images

Wang, Jinwang; Ding, Jian; Guo, Haowen; Cheng, Wensheng; Pan, Tongyan; Yang, Wen

doi:10.3390/rs11242930

Cited by 141 publications

(58 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second stage completes the three tasks of horizontal bounding box (HBB) prediction, oriented bounding box (OBB) prediction, and semantic segmentation. The HBB head predicts HBB representations of the bridges, while the OBB head firstly generates the instance masks and then converts them into OBB representations via the method proposed in Mask OBB [24]. The semantic head takes several fully convolutional layers to predict the water segmentation map in the image.…”

Section: Methodsmentioning

confidence: 99%

“…Ding et al [19] propose RoI Transformer, which introduces a rotational RoI Learner for RPN that significantly reduces the number of preset RoIs. Wang et al [24] take an instance-segmentation-based method to generate oriented bounding boxes from instance masks called Mask OBB. Han et al [20] address the severe misalignment between oriented anchor boxes and axis-aligned convolutional features by proposing a Feature Alignment Module for anchor generation.…”

Section: A Object Detection In Aerial Imagesmentioning

confidence: 99%

“…(2) Note cls as object classification cross entropy (CE) loss, reg as bounding box regression smooth 1 loss [12], OBB as oriented bounding box regression loss [24].…”

Section: Multi-task Loss Functionmentioning

confidence: 99%

See 2 more Smart Citations

Accurate Bridge Detection in Aerial Images With an Auxiliary Waterbody Extraction Task

Guo¹,

Zhang²,

Wang³

et al. 2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

Self Cite

View full text Add to dashboard Cite

Bridge detection in aerial images is to determine whether a given aerial image contains one or more bridges and locate them. However, the arbitrary orientations, extreme aspect ratios, and variable backgrounds pose great challenges for bridge detection and positioning. In this paper, we tackle these problems by combining the strengths of semantic-segmentation-based auxiliary supervision, waterbody constraint, and instance-switchingbased data augmentation. More precisely, we make three main contributions: (i) We propose an oriented bridge detection model with an auxiliary task of waterbody segmentation, which performs as guidance for bridge localization. The network is specifically designed in cascade style to handle the bridge detection and waterbody segmentation task end-to-end. (ii) We make use of the semantic features of waterbody as spatial attention to distinguish bridges from cluttered backgrounds, and then generate the waterbody segmentation map as the waterbody constraint, which introduces the prior knowledge of bridge distribution to refine the network predictions. (iii) We propose a background consistent instance switching method for online data augmentation to further improve the robustness of bridge detection. To verify the effectiveness of the proposed method, we introduce a dataset named BridgeDetV1 containing 5,000 well-annotated images with two kinds of bridge representations, i.e., the horizontal bounding box and the oriented bounding box. Extensive experiments demonstrate that our approach outperforms the state-of-the-art methods on this challenging benchmark. Dataset and code are available at https://github.com/whughw/BridgeDet.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: A Object Detection In Aerial Imagesmentioning

confidence: 99%

See 1 more Smart Citation

Accurate Bridge Detection in Aerial Images With an Auxiliary Waterbody Extraction Task

Guo¹,

Zhang²,

Wang³

et al. 2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

Self Cite

View full text Add to dashboard Cite

show abstract

“…Therefore, the HBB-based model may cause severe overlap and noise. In subsequent work, an oriented bounding box (OBB) was used to process rotating remote sensing targets [32][33][34][35][36][37][38][39][40], enabling more accurate target capture and introducing considerably less background noise. Feng et al [32] proposed a robust Student's t-distribution-aided one-stage orientation detector.…”

Section: Introductionmentioning

confidence: 99%

Improved YOLO Network for Free-Angle Remote Sensing Target Detection

et al. 2021

View full text Add to dashboard Cite

Despite significant progress in object detection tasks, remote sensing image target detection is still challenging owing to complex backgrounds, large differences in target sizes, and uneven distribution of rotating objects. In this study, we consider model accuracy, inference speed, and detection of objects at any angle. We also propose a RepVGG-YOLO network using an improved RepVGG model as the backbone feature extraction network, which performs the initial feature extraction from the input image and considers network training accuracy and inference speed. We use an improved feature pyramid network (FPN) and path aggregation network (PANet) to reprocess feature output by the backbone network. The FPN and PANet module integrates feature maps of different layers, combines context information on multiple scales, accumulates multiple features, and strengthens feature information extraction. Finally, to maximize the detection accuracy of objects of all sizes, we use four target detection scales at the network output to enhance feature extraction from small remote sensing target pixels. To solve the angle problem of any object, we improved the loss function for classification using circular smooth label technology, turning the angle regression problem into a classification problem, and increasing the detection accuracy of objects at any angle. We conducted experiments on two public datasets, DOTA and HRSC2016. Our results show the proposed method performs better than previous methods.

show abstract

“…One is box-of-free feature extraction [13][14][15], in which target detection is accomplished by embedding a cosine function or embedding a class of clusters in pixels. The other is based on frame-based feature extraction, but this method of embedding clusters has two major disadvantages in the extraction process [16]. One is that the global information of the picture cannot be fully considered, and the other is that the embedded information is mainly a cosine function, so there are many restrictions before embedding, and this method must be limited in the use process.…”

Section: Introductionmentioning

confidence: 99%

PGNet: Pipeline Guidance for Human Key-Point Detection

Hong

Liu

et al. 2020

Entropy

View full text Add to dashboard Cite

Human key-point detection is a challenging research field in computer vision. Convolutional neural models limit the number of parameters and mine the local structure, and have made great progress in significant target detection and key-point detection. However, the features extracted by shallow layers mainly contain a lack of semantic information, while the features extracted by deep layers contain rich semantic information but a lack of spatial information that results in information imbalance and feature extraction imbalance. With the complexity of the network structure and the increasing amount of computation, the balance between the time of communication and the time of calculation highlights the importance. Based on the improvement of hardware equipment, network operation time is greatly improved by optimizing the network structure and data operation methods. However, as the network structure becomes deeper and deeper, the communication consumption between networks also increases, and network computing capacity is optimized. In addition, communication overhead is also the focus of recent attention. We propose a novel network structure PGNet, which contains three parts: pipeline guidance strategy (PGS); Cross-Distance-IoU Loss (CIoU); and Cascaded Fusion Feature Model (CFFM).

show abstract

Mask OBB: A Semantic Attention-Based Mask Oriented Bounding Box Representation for Multi-Category Object Detection in Aerial Images

Cited by 141 publications

References 57 publications

Accurate Bridge Detection in Aerial Images With an Auxiliary Waterbody Extraction Task

Accurate Bridge Detection in Aerial Images With an Auxiliary Waterbody Extraction Task

Improved YOLO Network for Free-Angle Remote Sensing Target Detection

PGNet: Pipeline Guidance for Human Key-Point Detection

Contact Info

Product

Resources

About