Structure-Aware Face Clustering on a Large-Scale Graph with 10<sup>7</sup>Nodes

Shen, Shuai; Li, Wanhua; Zhu, Zheng; Huang, Guan; Du, Dalong; Lu, Jiwen; Zhou, Jie

doi:10.1109/cvpr46437.2021.00897

Cited by 17 publications

(34 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Face clustering has been thoroughly investigated in recent years. Significant performance improvements (Wang et al, 2019b;Guo et al, 2020;Shen et al, 2021) have been obtained with Graph Convolutional Networks due to their powerful feature propagation capacity. The representative DA-Net (Guo et al, 2020) and STAR-FC (Shen et al, 2021) use GCNs to learn enhanced feature embedding by vertices or edges classification tasks to assist clustering.…”

Section: Introductionmentioning

confidence: 99%

“…Significant performance improvements (Wang et al, 2019b;Guo et al, 2020;Shen et al, 2021) have been obtained with Graph Convolutional Networks due to their powerful feature propagation capacity. The representative DA-Net (Guo et al, 2020) and STAR-FC (Shen et al, 2021) use GCNs to learn enhanced feature embedding by vertices or edges classification tasks to assist clustering. However, the main problem restricting the power of existing GCN-based face clustering algorithms is the existence of noise edges in the face graphs.…”

Section: Introductionmentioning

confidence: 99%

“…Unlike common graph datasets such as Citeseer, Cora and Pubmed with explicit link relation as edges (Kipf & Welling, 2017), face images do not contain explicit structural information, but only deep features extracted from a trained CNN model. Therefore, face images are treated as vertices, and the edges between face images are usually constructed based on the kNN (Cover & Hart, 1967) relations when building the graph: Each face serves as a probe to retrieve its k nearest neighbours by deep features (Wang et al, 2019b;Guo et al, 2020;Shen et al, 2021). The kNN relations are not always reliable because deep The existing "One-size-fits-all" solution using a fixed number of neighbours for each vertex introduces many noise edges.…”

Section: Introductionmentioning

confidence: 99%

“…The noise edges problem is common in face clustering but has received little research attention. For example, the graph used in Shen et al, 2021) contains about 38.23% noise edges in testing. The noise edges will propagate noisy information between vertices, hurting their features when aggregation, and thus resulting in inferior performance.…”

Section: Introductionmentioning

confidence: 99%

“…In Figure 1 (b), the triangular vertex v 1 is connected with three vertices of different classes, and will it be polluted by them with messages passing in the graph. Therefore, GCN-based linkage prediction cannot effectively address the noise edges problem in related works (Wang et al, 2019b;Shen et al, 2021).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Ada-NETS: Face Clustering via Adaptive Neighbour Discovery in the Structure Space

Wang¹,

Zhang²,

Zhang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Face clustering has attracted rising research interest recently to take advantage of massive amounts of face images on the web. State-of-the-art performance has been achieved by Graph Convolutional Networks (GCN) due to their powerful representation capacity. However, existing GCN-based methods build face graphs mainly according to kNN relations in the feature space, which may lead to a lot of noise edges connecting two faces of different classes. The face features will be polluted when messages pass along these noise edges, thus degrading the performance of GCNs. In this paper, a novel algorithm named Ada-NETS is proposed to cluster faces by constructing clean graphs for GCNs. In Ada-NETS, each face is transformed to a new structure space, obtaining robust features by considering face features of the neighbour images. Then, an adaptive neighbour discovery strategy is proposed to determine a proper number of edges connecting to each face image. It significantly reduces the noise edges while maintaining the good ones to build a graph with clean yet rich edges for GCNs to cluster faces. Experiments on multiple public clustering datasets show that Ada-NETS significantly outperforms current state-of-the-art methods, proving its superiority and generalization.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Ada-NETS: Face Clustering via Adaptive Neighbour Discovery in the Structure Space

Wang¹,

Zhang²,

Zhang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Learning to Cluster Faces with Mixed Face Quality

Pan¹,

Guo²,

Yang³

et al. 2022

Pattern Recognition and Computer Vision

View full text Add to dashboard Cite

Large-Scale Clustering on 100 M-Scale Datasets Using a Single T4 GPU via Recall KNN and Subgraph Segmentation

Liu,

Jiang,

Liu

et al. 2024

Neural Process Lett

View full text Add to dashboard Cite

Despite the promising progress that has been made, large-scale clustering tasks still face various challenges: (i) high time and space complexity in K-nearest neighbors (KNN), which is often overlooked by most methods, and (ii) low recall rate caused by simply splitting the dataset. In this paper, we propose a novel framework for large-scale clustering tasks named large-scale clustering via recall KNN and subgraph segmentation (LS-RKSS) to perform faster clustering with guaranteed clustering performance, which embraces the ability of handling large-scale data up to 100 million using a single T4 GPU with less than 10% of the running time. We propose recall KNN (RKNN) and subgraph segmentation (SS) to effectively address the primary challenges in large-scale clustering tasks. Firstly, the recall KNN is proposed to perform efficient similarity search among dense vectors with lower time and space complexity compared to traditional exact search methods of KNN. Then, the subgraph segmentation is proposed to split the whole dataset into multiple subgraphs based on the recall KNN. Given the recall rate of RKNN based on traditional exact search methods, it is theoretically proved that dividing the dataset into multiple subgraphs using recall KNN and subgraph segmentation is a more reasonable and effective approach. Finally, clusters are generated independently on each subgraph, and the final clustering result is obtained by combining the results of all subgraphs. Extensive experiments demonstrate that LS-RKSS outperforms previous large-scale clustering methods in both effectiveness and efficiency.

show abstract

Structure-Aware Face Clustering on a Large-Scale Graph with 10⁷Nodes

Cited by 17 publications

References 35 publications

Ada-NETS: Face Clustering via Adaptive Neighbour Discovery in the Structure Space

Ada-NETS: Face Clustering via Adaptive Neighbour Discovery in the Structure Space

Learning to Cluster Faces with Mixed Face Quality

Large-Scale Clustering on 100 M-Scale Datasets Using a Single T4 GPU via Recall KNN and Subgraph Segmentation

Contact Info

Product

Resources

About

Structure-Aware Face Clustering on a Large-Scale Graph with 107Nodes

Cited by 17 publications

References 35 publications

Ada-NETS: Face Clustering via Adaptive Neighbour Discovery in the Structure Space

Ada-NETS: Face Clustering via Adaptive Neighbour Discovery in the Structure Space

Learning to Cluster Faces with Mixed Face Quality

Large-Scale Clustering on 100 M-Scale Datasets Using a Single T4 GPU via Recall KNN and Subgraph Segmentation

Contact Info

Product

Resources

About

Structure-Aware Face Clustering on a Large-Scale Graph with 10⁷Nodes