Vision language models are used in zero-shot mode to infer vehicle make/model/generation and accurate 3D dimensions from image crops, improving label quality and reducing manual effort especially under occlusion.
Argoverse 2: Next generation datasets for self-driving perception and forecasting
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Improving 3D Labeling in Self-Driving by Inferring Vehicle Information using Vision Language Models
Vision language models are used in zero-shot mode to infer vehicle make/model/generation and accurate 3D dimensions from image crops, improving label quality and reducing manual effort especially under occlusion.