Turn image captions into labeled bounding boxes.
What it does
Generates training annotations that link phrases in image captions to pixel-space bounding boxes.
When to use it
Use when grounding captions to bounding boxes, generating phrase-grounded annotations, or automatically labeling images for grounding.
How to use it
Give it images and a caption JSONL file; it asks about endpoint access and workflow choices, then produces annotation files.
What you provide
- Your files
Uses
Latest review
What you need · 7
Requires Docker to run the containerized pipeline.
Requires NVIDIA Container Toolkit alongside Docker.
Requires the auto_label CLI inside the TAO Toolkit container, specified as nvcr.io/nvidia/tao/tao-toolkit:7.3.0-pyt.
Access · 1
This skill
VLM endpoint
Write
Sends images and captions for annotation