Learning to Cluster Faces via Confidence and Connectivity Estimation

Yang, Lei; Chen, Dapeng; Zhan, Xiaohang; Zhao, Rui; Loy, Chen Change; Lin, Dahua

doi:10.1109/cvpr42600.2020.01338

Cited by 73 publications

(101 citation statements)

References 28 publications

Supporting

Mentioning

101

Contrasting

Order By: Relevance

“…A GCN-based detection and segmentation module was proposed to complete the face clustering task. Yang et al [10] used GCN to infer the confidence of nodes and the connectivity of edges to complete clustering. Guo et al [11] fused GCN and long short-term memory (LSTM) to obtain embedded face data based on density and then used traditional algorithms to achieve a good effect.…”

Section: A Face Clusteringmentioning

confidence: 99%

“…The drawbacks of these algorithms limit their application in face clustering problems with complex distributions of facial representations. To address this constellation of real problems, recent studies have shown that utilizing graph convolutional networks (GCNs) and supervised information [7], [8], [10], [13] can enhance the characteristics of face clustering. GCNs learn cluster patterns rather than completing a cluster.…”

mentioning

confidence: 99%

“…Although these graph-convolution-based supervised methods can effectively deal with complex clustering patterns, they also have some problems. First, they construct numerous overlapping subgraphs [7], [10]. The subgraphs are overly redundant, which increases the consumption of computing resources and severely limits their efficiency.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Confidence-Based Simple Graph Convolutional Networks for Face Clustering

2022

View full text Add to dashboard Cite

Face clustering is an effective method for taking advantage of unlabeled face data. Recent studies use graph convolutional networks (GCNs) to learn feature embeddings from the neighborhood information between face images. However, most of the face clustering methods require numerous overlapping subgraphs to characterize the local structure around the nodes, which causes significant redundancy. Moreover, the nonlinearity of the GCN itself increases the calculation complexity, which further reduces the model's training efficiency. In this study, we propose a lightweight clustering framework, the confidencebased simple graph convolutional network (CSGCN), for face clustering, which achieves more accurate clustering results and significantly improves the efficiency of GCN-based face clustering. Specifically, CSGCN does not construct any subgraphs but convolves the entire graph as a whole and also removes the nonlinearity of the convolution in the graph convolution module, which further reduces the computational complexity. Subsequently, an effective new confidence score is constructed to better characterize the embedded features and to ensure that the subsequent clustering still maintains a high accuracy rate under the aforementioned model simplification. In addition, while most of the existing GCN-based methods are actually supervised, we construct an unsupervised confidence to make it more suitable for clustering tasks. Extensive experiments with MS-Celeb-1M, YouTube-Faces and DeepFashion datasets show that our method not only improves the clustering accuracy but also significantly reduces the execution time, whether in supervised or unsupervised models.

show abstract

Section: A Face Clusteringmentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

Confidence-Based Simple Graph Convolutional Networks for Face Clustering

2022

View full text Add to dashboard Cite

show abstract

“…We assess the performance on both person clustering and relationship discovery. For person clustering, we evaluate it with two widely used metrics [18,19], namely Identification and F-score. We adopt top-1 identification hit rate, which is to rank the top-1 image from the 100 gallery images and compute the top-1 hit rate.…”

Section: Metricsmentioning

confidence: 99%

“…The feature extraction is performed by a ResNet-50 and it can be replaced by a lightweight backbone in practice. The complexity of person clustering is similar to the proposal generation procedure of [18] and it has a more efficient alternative [19] for real-world applications. The relationship discovery only adds a little computation compared to person clustering, as the entire graph has been clustered into some large groups in the previous stage.…”

Section: Metricsmentioning

confidence: 99%

Cohabitation Discovery via Spatial and Temporal Clustering

Liu¹

2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Recent years have witnessed the rapid development of deep learning in various aspects, such as image classification, face recognition, and object detection. Yet, these approaches often focus on a single entity. The relationship between different entities is remained to be explored. Cohabitation is a kind of important relationship. In a scenario of residential entries, knowing the relationship of cohabitation could a) prevent tailgaters; b) identify unregistered strangers, and especially c) prevent the disease from spreading during the Cov-19 period. In this paper, we propose a method combining computer vision with graph algorithms to discover the cohabitation relationships as long-term and regular co-occurrence. We demonstrate the method beneficial to both industrial and technical aspects.

show abstract