Zero-Label Driving Scenario Complexity Detection via Joint Embedding Predictive Architecture
Quick Answer
The study introduces a Joint Embedding Predictive Architecture (JEPA) that autonomously detects driving scenario complexity without labels, achieving significant differentiation in complexity scores for various scenarios.
Quick Take
The study introduces a Joint Embedding Predictive Architecture (JEPA) that autonomously detects driving scenario complexity without labels, achieving significant differentiation in complexity scores for various scenarios. The model demonstrated an Average Precision of 0.512 in anomaly detection, outperforming a baseline of 0.436, highlighting its potential in identifying critical driving situations.
Key Points
- JEPA trained on nuPlan mini dataset detects scenario complexity without any labels.
- Higher complexity scores assigned to unprotected turns and pedestrian interactions.
- Lower scores for simpler scenarios like lane-following and stationary traffic.
- Achieved Average Precision of 0.512 in anomaly detection tasks.
- Four ablation experiments validated the model's effectiveness.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Identifying complex and safety-critical driving scenarios in large unlabelled datasets is an important but expensive problem. Existing approaches rely on human annotators, supervised classifiers, or carefully engineered rule sets, all of which require substantial prior knowledge about what constitutes a difficult scenario. We ask whether a model can discover scenario complexity on its own, with no labels at any stage. We train a minimal Joint Embedding Predictive Architecture (JEPA) on structured agent state data from the nuPlan mini dataset and use the temporal prediction error as a zero-shot complexity score. Without access to any ground-truth labels during training or evaluation setup, the model assigns significantly higher scores to scenarios involving unprotected turns, crosswalk interactions, and pedestrian proximity, and significantly lower scores to lane-following and stationary-traffic scenarios. We validate this finding through four ablation experiments that isolate the source of the signal, and through a downstream anomaly detection evaluation that achieves Average Precision of 0.512 against a 0.436 chance baseline. The results show that temporal prediction error in a self-supervised latent world model is a practical proxy for driving scenario complexity.
| Comments: | 12 pages, 6 figures |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2606.28383 [cs.CV] |
| (or arXiv:2606.28383v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2606.28383 arXiv-issued DOI via DataCite |
Submission history
From: Santosh Jaiswal [view email]
[v1]
Sun, 21 Jun 2026 05:26:58 UTC (877 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →LLM-Guided ANN Index Optimization for Human-Object Interaction Retrieval
A phase-aware LLM agent optimizes human-object interaction retrieval, outperforming Optuna TPE by 33.3% and VDTuner by 34.2% on the HICO-DET benchmark. This method enhances throughput by 15.3x over UniIR and demonstrates strong transferability across vector database management systems.