Turn labeled images into grounded descriptions.
What it does
Generates grounded image annotations from KITTI bounding boxes, including object descriptions, scene captions, and optionally verified referring expressions.
When to use it
Use when generating referring-expression annotations, region descriptions, or grouped grounding phrases from images with KITTI labels, including optional verification.
How to use it
Give it images with KITTI labels or existing annotations; it asks about the VLM endpoint, workflow steps, and output format, then generates annotation files.
What you provide
- Your files
Uses
What you need · 9
Docker is required to run the container-based pipeline.
The compatibility declaration requires NVIDIA Container Toolkit.
Requires the specified TAO Toolkit container image, in which the auto_label CLI runs.
Access · 1
This skill
VLM endpoint
Sends images to your VLM