Remember Intentions: Retrospective-Memory-based Trajectory Prediction

Xu, Chenxin; Mao, Weibo; Zhang, Wenjun; Chen, Siheng

doi:10.1109/cvpr52688.2022.00638

Cited by 50 publications

(18 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also report the performance of different methods on SDD in Table 8. It shows that our model performs a little bit worse than the results of YNet [19] and MemoNet [32]. Nevertheless, our method performs better than previous baselines (such as PECNet [18], MANTRA [32]).…”

Section: B2 Experimental Resultsmentioning

confidence: 67%

“…It shows that our model performs a little bit worse than the results of YNet [19] and MemoNet [32]. Nevertheless, our method performs better than previous baselines (such as PECNet [18], MANTRA [32]). Consequently, our method can achieve reasonable performance in bird-eye-view scenario.…”

Section: B2 Experimental Resultsmentioning

confidence: 67%

“…In our work, we integrate the information from the scene history in a more comprehensive way and propose dedicated modules to handle different types of information, respectively. The main differences between our method and the previous works, especially memory-based methods [15,20,21,32] and clustering-based methods [28,29,36], are as follows: i) MANTRA [20] and MemoNet [32] consider historical individual trajectories, while our proposed SHENet focuses on historical group trajectories, which is more generalizable in different scenarios; In [15], historical trajectories are not stored for search.…”

Section: Related Workmentioning

confidence: 99%

“…Most attempts [16,19,27,35] explore only individual-surroundings interaction, where tremendous efforts are spent on modeling the individual trajectory, the semantic information of environment information, and the relationship between them. Although MANTRA [20] models the similarity using the encoders trained in a reconstruction manner and MemoNet [32] simplifies the similarity by storing the intention of historical trajectories, they both perform similarity computation on instance level, instead of group level, making it sensitive to the capability of the trained encoder. Based on the above analysis, we propose a simple yet effective framework, Scene History Excavating Network (SHENet), to leverage HGT and ISI jointly for HTP.…”

Section: Introductionmentioning

confidence: 99%

“…We also conduct the experiments on ETH/UCY without using the video data, shown in Table7. Since MANTRA didn't conduct experiments on ETH/UCY, we use the results of MANTRA reported in the work[32].…”

mentioning

confidence: 99%

See 4 more Smart Citations

Forecasting Human Trajectory from Scene History

Meng¹,

Wu²,

Chen³

et al. 2022

Preprint

View full text Add to dashboard Cite

Predicting the future trajectory of a person remains a challenging problem, due to randomness and subjectivity of human movement. However, the moving patterns of human in a constrained scenario typically conform to a limited number of regularities to a certain extent, because of the scenario restrictions (e.g., floor plan, roads, and obstacles) and person-person or person-object interactivity. Thus, an individual person in this scenario should follow one of the regularities as well. In other words, a person's subsequent trajectory has likely been traveled by others. Based on this hypothesis, we propose to forecast a person's future trajectory by learning from the implicit scene regularities. We call the regularities, inherently derived from the past dynamics of the people and the environment in the scene, scene history. We categorize scene history information into two types: historical group trajectory and individual-surroundings interaction. To exploit these two types of information for trajectory prediction, we propose a novel framework Scene History Excavating Network (SHENet), where the scene history is leveraged in a simple yet effective approach. In particular, we design two components: the group trajectory bank module to extract representative group trajectories as the candidate for future path, and the cross-modal interaction module to model the interaction between individual past trajectory and its surroundings for trajectory refinement. In addition, to mitigate the uncertainty in ground-truth trajectory, caused by the aforementioned randomness and subjectivity of human movement, we propose to include smoothness into the training process and evaluation metrics. We conduct extensive evaluations to validate the efficacy of our proposed framework on ETH, UCY, as well as a new, challenging benchmark dataset PAV, demonstrating superior performance compared to state-of-the-art methods. Code is available at: https://github.com/MaKaRuiNah/SHENet

show abstract