PolyWorld: Polygonal Building Extraction with Graph Neural Networks in Satellite Images

Zorzi, Stefano; Bazrafkan, Shabab; Habenschuss, Stefan; Fraundorfer, Friedrich

doi:10.1109/cvpr52688.2022.00189

Cited by 64 publications

(30 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our experiments, we select PolyMapper [24], Framefield [29], and PolyWorld [26]) for comparison, which are recently proposed state-of-the-art (SOTA) methods for polygonal building extraction. Besides, we compare our method with classical instance segmentation methods, including Mask RCNN [36] and PANet [62] following the recent SOTA methods.…”

Section: Resultsmentioning

confidence: 99%

“…Recently, PolyWorld [26] directly predicts all vertices and then generates a connection matrix to find the vertices of one building and the order of building vertices. Since Poly-World produces serialized vertices in a bottom-up pathway, missing or error vertices will influence the connection matrix learning, leading to self-intersection or non-closed polygons.…”

Section: B Serialized Vertices Based Building Mappingmentioning

confidence: 99%

“…Polygonal metric. The polygonal metric considers the geometric properties of the extracted buildings, so we adopt three indicators to evaluate the extracted polygonal buildings, which consist of the N ratio [26], the complexity aware IoU (C-IoU) [26], and the Max Tangent Angle Error (MTA) [29]. The N ratio measures the simplicity of polygonal buildings by calculating the ratio between the predicted vertex number and the ground truth, which is defined as:…”

Section: B Evaluation Metricsmentioning

confidence: 99%

“…Recently, dominant building mapping approaches [24]- [26] represent building extraction as polygonal building vertex prediction and directly predict building vertices to produce vector polygons of buildings. These methods can be categorized as follows: (1) Predict serialized vertices clockwise or counterclockwise from the building feature map of the candidate building region aligned from remote sensing image features.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Multi-Layer Fault-Tolerant Protection Strategies for Hybrid Distribution Transformers Integrated Photovoltaic Systems

Zhang

Liu

Wang

et al. 2023

IEEE Trans. on Ind. Applicat.

View full text Add to dashboard Cite

Deep learning-based methods have been extensively explored for automatic building mapping from high-resolution remote sensing images over recent years. While most building mapping models produce vector polygons of buildings for geographic and mapping systems, dominant methods typically decompose polygonal building extraction in some sub-problems, including segmentation, polygonization, and regularization, leading to complex inference procedures, low accuracy, and poor generalization. In this paper, we propose a simple and novel building mapping method with Hierarchical Transformers, called HiT, improving polygonal building mapping quality from high-resolution remote sensing images. HiT builds on a two-stage detection architecture by adding a polygon head parallel to classification and bounding box regression heads. HiT simultaneously outputs building bounding boxes and vector polygons, which is fully end-to-end trainable. The polygon head formulates a building polygon as serialized vertices with the bidirectional characteristic, a simple and elegant polygon representation avoiding the start or end vertex hypothesis. Under this new perspective, the polygon head adopts a transformer encoder-decoder architecture to predict serialized vertices supervised by the designed bidirectional polygon loss. Furthermore, a hierarchical attention mechanism combined with convolution operation is introduced in the encoder of the polygon head, providing more geometric structures of building polygons at vertex and edge levels. Comprehensive experiments on two benchmarks (the CrowdAI and Inria datasets) demonstrate that our method achieves a new state-of-the-art in terms of instance segmentation and polygonal metrics compared with state-of-theart methods. Moreover, qualitative results verify the superiority and effectiveness of our model under complex scenes.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: B Serialized Vertices Based Building Mappingmentioning

confidence: 99%

Section: B Evaluation Metricsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Multi-Layer Fault-Tolerant Protection Strategies for Hybrid Distribution Transformers Integrated Photovoltaic Systems

Zhang

Liu

Wang

et al. 2023

IEEE Trans. on Ind. Applicat.

View full text Add to dashboard Cite

show abstract

“…This concept forms the core of DCNs (Dai et al, 2017; Zhu et al, 2019). The strategy of learned pixel‐wise deformations has been widely used (Peng et al, 2020; Wei et al, 2023; Zorzi et al, 2021) and is applied in this study for adaptive image matching.…”

Section: Related Workmentioning

confidence: 99%

Adaptive region aggregation for multi‐view stereo matching using deformable convolutional networks

Hu,

Su,

Mao

et al. 2023

The Photogrammetric Record

View full text Add to dashboard Cite

Deep‐learning methods have demonstrated promising performance in multi‐view stereo (MVS) applications. However, it remains challenging to apply a geometrical prior on the adaptive matching windows to achieve efficient three‐dimensional reconstruction. To address this problem, this paper proposes a learnable adaptive region aggregation method based on deformable convolutional networks (DCNs), which is integrated into the feature extraction workflow for MVSNet method that uses coarse‐to‐fine structure. Following the conventional pipeline of MVSNet, a DCN is used to densely estimate and apply transformations in our feature extractor, which is composed of a deformable feature pyramid network (DFPN). Furthermore, we introduce a dedicated offset regulariser to promote the convergence of the learnable offsets of the DCN. The effectiveness of the proposed DFPN is validated through quantitative and qualitative evaluations on the BlendedMVS and Tanks and Temples benchmark datasets within a cross‐dataset evaluation setting.

show abstract

A shoreline extraction method based on dual-loop network framework

Li,

Cao,

et al. 2024

Vis Comput

View full text Add to dashboard Cite

PolyWorld: Polygonal Building Extraction with Graph Neural Networks in Satellite Images

Cited by 64 publications

References 30 publications

Multi-Layer Fault-Tolerant Protection Strategies for Hybrid Distribution Transformers Integrated Photovoltaic Systems

Multi-Layer Fault-Tolerant Protection Strategies for Hybrid Distribution Transformers Integrated Photovoltaic Systems

Adaptive region aggregation for multi‐view stereo matching using deformable convolutional networks

A shoreline extraction method based on dual-loop network framework

Contact Info

Product

Resources

About