VLA (vision-language-action) labelling
Ground truth for the language and action layer of autonomy — turning driving and robotics scenes into supervision for vision-language-action models, from scene meaning to intent and decisions.
- Scene captioning and natural-language descriptions
- Referring expressions and object grounding
- Driving-action, intent, and decision annotation
- Question-and-answer pairs for scene understanding
- Instruction-to-trajectory and decision labeling
- Camera–LiDAR multimodal grounding