AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

Social-SSL: Self-Supervised Cross-Sequence Representation Learning Based on Transformers for Multi-agent Trajectory Prediction

Li-Wu Tsao, Yan-Kai Wang, Hao-Siang Lin, Hong-Han Shuai, Lai-Kuan Wong, Wen-Huang Cheng

ECCV 2022 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Interactive multi-agent trajectory prediction · 31 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add closeness pretext task (ablation)Social-SSL · SDDADE/FDE: 7.64/14.27 -> 6.63/12.23Improved
add interaction pretext task (ablation)Social-SSL · SDDADE/FDE: 11.27/20.94 -> 6.63/12.23Improved
add segment embeddings (ablation)Social-SSL · SDDADE/FDE: 9.23/17.75 -> 6.63/12.23Improved
add segment embeddings (ablation)Ours-SP · SDDADE/FDE: 17.65/31.62 -> 11.96/22.36Improved
add motion-related (masked sequence) pretext (ablation)Social-SSL social pretexts · SDDADE/FDE: 26.81/46.72 -> 6.63/12.23Improved
add social-related pretexts (ablation)Social-SSL · SDDADE/FDE: 11.96/22.36 -> 6.63/12.23Improved
extend single-agent Transformer to cross-sequence input without pre-training (ablation)Transformer baseline · SDDADE/FDE: 16.88/31.50 -> 26.46/48.45Hurt
add social pretexts without motion pretext (ablation)cross-agent baseline · SDDADE/FDE: 26.46/48.45 -> 26.81/46.72No clear effect
combine sparse and dense closeness settingsSocial-SSL · ETH/UCY avgADE/FDE: 0.45/0.86 -> 0.44/0.85No clear effect
use 100% instead of 10% training dataSocial-SSL · ETH/UCY avgADE/FDE: 0.59/1.14 -> 0.44/0.85Improved
use Social-SSL pre-training instead of Social-STGCNNtrajectory predictor · ETH/UCY avg, 10% dataADE/FDE: 1.48/2.35 -> 0.59/1.14Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy