Papers › THOMAS: Trajectory Heatmap Output with learned Multi-Agent Sampling

THOMAS: Trajectory Heatmap Output with learned Multi-Agent Sampling

13 Oct 2021ICLR 2022 4arXiv:2110.06607archive 2025-07-28

Thomas Gilles, Stefano Sabatini, Dzmitry Tsishkou, Bogdan Stanciulescu, Fabien Moutarde

In this paper, we propose THOMAS, a joint multi-agent trajectory prediction framework allowing for an efficient and consistent prediction of multi-agent multi-modal trajectories. We present a unified model architecture for simultaneous agent future heatmap estimation, in which we leverage hierarchical and sparse image generation for fast and memory-efficient inference. We propose a learnable trajectory recombination model that takes as input a set of predicted trajectories for each agent and outputs its consistent reordered recombination. This recombination module is able to realign the initially independent modalities so that they do no collide and are coherent with each other. We report our results on the Interaction multi-agent prediction challenge and rank 1ˢᵗ on the online test leaderboard.

PaperPDFConference PDF

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image GenerationPredictionTrajectory Prediction

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Motion Forecasting Argoverse CVPR 2020 THOMAS DAC (K=6) 0.9781 #67 of 299 Archive leaderboard report
Motion Forecasting Argoverse CVPR 2020 THOMAS MR (K=1) 0.5613 #67 of 299 Archive leaderboard report
Motion Forecasting Argoverse CVPR 2020 THOMAS MR (K=6) 0.1038 #67 of 299 Archive leaderboard report
Motion Forecasting Argoverse CVPR 2020 THOMAS brier-minFDE (K=6) 1.9736 #67 of 299 Archive leaderboard report
Motion Forecasting Argoverse CVPR 2020 THOMAS minADE (K=1) 1.6686 #67 of 299 Archive leaderboard report
Motion Forecasting Argoverse CVPR 2020 THOMAS minADE (K=6) 0.9423 #67 of 299 Archive leaderboard report
Motion Forecasting Argoverse CVPR 2020 THOMAS minFDE (K=1) 3.593 #67 of 299 Archive leaderboard report
Motion Forecasting Argoverse CVPR 2020 THOMAS minFDE (K=6) 1.4388 #67 of 299 Archive leaderboard report
Trajectory Prediction nuScenes THOMAS MinADE_10 1.04 #8 of 34 Archive leaderboard report
Trajectory Prediction nuScenes THOMAS MinADE_5 1.33 #8 of 34 Archive leaderboard report
Trajectory Prediction nuScenes THOMAS MinFDE_1 6.71 #8 of 34 Archive leaderboard report
Trajectory Prediction nuScenes THOMAS MissRateTopK_2_10 0.42 #8 of 34 Archive leaderboard report
Trajectory Prediction nuScenes THOMAS MissRateTopK_2_5 0.55 #8 of 34 Archive leaderboard report
Trajectory Prediction nuScenes THOMAS OffRoadRate 0.03 #8 of 34 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

HeatmapTest

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections