Self-supervised Social Relation Representation for Human Group Detection

Li, Jiacheng; Han, Ruize; Yan, Haomin; Qian, Zekun; Feng, Wei; Wang, Song

doi:10.1007/978-3-031-19833-5_9

Cited by 4 publications

(18 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, the occlusion of appearance features of different persons in long-duration crowd videos was not properly considered, leading to inaccurate grouping results. Recently, Li et al [15] proposed a two-stage method that pre-trains the model on unsupervised tasks before fine-tuning for group detection. While achieving promising performance, this approach neglects frequent occlusions in crowd videos and relies on a gated recurrent unit model for temporal information aggregation, which may limit its effectiveness.…”

Section: B Dynamic Methodsmentioning

confidence: 99%

“…Besides, in group detection, the temporal information e.g., trajectories, is not just a sequence, it is also the important location information for distinguishing the group results. LSGD [24] and S3R2 [15] handle temporal information and spatial information respectively, which can result in suboptimal performance.…”

Section: Spatio-temporal Methodsmentioning

confidence: 99%

“…Given a video of a large-scale multi-person scene, our method can detect reasonable groups, despite the large number of people and frequent occlusions in the scene. While the state-of-the-art method [15] fails to predict right results. Different groups are shown with different color boxes.…”

Section: Introductionmentioning

confidence: 98%

“…It is worth noting that 3D ConvNet faces similar challenges as I3D, and LSTM fails to incorporate spatial information from trajectories. Recently, Li et al [15] proposed a novel group detection method by pre-training the model using a self-supervised method, which produced promising results. However, this method also disregarded the occlusion problem and processed spatial and temporal information separately, which means that the spatial information and the temporal information are processed in sequence and not synchronized [22], limiting its performance.…”

Section: Introductionmentioning

confidence: 99%

“…We evaluate our method on both small-scale scene datasets and large-scale scene datasets, and the experimental results demonstrate its superiority over state-of-the-art methods. Figure 1 illustrates an example of the grouping results compared our method with the state-of-the-art method, S3R2 [15]. To facilitate further research, we will release the code for GroupTransformer for academic purposes.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Toward Grouping in Large Scenes With Occlusion-Aware Spatio–Temporal Transformers

Zhang,

Gu,

Lai

et al. 2024

IEEE Trans. Circuits Syst. Video Technol.

View full text Add to dashboard Cite

Group detection, especially for large-scale scenes, has many potential applications for public safety and smart cities. Existing methods fail to cope with frequent occlusions in large-scale scenes with multiple persons, and are difficult to effectively utilize spatio-temporal information. In this paper, we propose an end-to-end framework, GroupTransformer, for group detection in large-scale scenes. To deal with the frequent occlusions caused by multiple people, we design an occlusion encoder to detect and suppress severely occluded person crops. To explore the potential spatio-temporal relationship, we propose spatio-temporal transformers to simultaneously extract trajectory information and fuse inter-person features in a hierarchical manner. Experimental results on both large-scale and small-scale scenes demonstrate that our method achieves better performance compared with state-of-the-art methods. On large-scale scenes, our method significantly boosts the performance in terms of precision and F1 score by more than 10%. On small-scale scenes, our method still improves the performance of F1 score by more than 5%. We will release the code for research purposes.

show abstract

Section: B Dynamic Methodsmentioning

confidence: 99%

Section: Spatio-temporal Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 98%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations