vlm-visualizer

Overlay VLM bounding box annotations from JSON onto images using Python and Pillow.

Updated Mar 6, 2026

Search Tags:#json #segmentation #pillow #visualize #bounding-box #image-annotation

Ready to run

npx skills add https://github.com/ghjghjghkimo/skills --skill vlm-visualizer

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: vlm-visualizer
Source: https://github.com/ghjghjghkimo/skills/tree/main/.gemini/skills/vlm-visualizer
Command: npx skills add https://github.com/ghjghjghkimo/skills --skill vlm-visualizer

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires Pillow, and includes scripts (resource) components.

What problem does it solve?

Generates visual verification by overlaying VLM bounding box JSON data onto the original image, helping validate segmentation results and produce clear visual reports.

Core Features & Use Cases

Overlay bounding boxes from a VLM JSON file on the source image to create annotated visuals.
Reads a panels array where each item provides a label and box_2d coordinates, enabling quick QA checks.
Use Case: produce visual reports for segmentation verification in CV pipelines and presentations.

Quick Start

Run the visualization script to overlay VLM bounding boxes on an image using your JSON data.

Frequently Asked Questions about vlm-visualizer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I overlay bounding box JSON data onto an image for verification?▼

To overlay bounding box JSON data, you need an image paired with a JSON file containing a panels array of objects with label and box_2d coordinates, which are converted to pixel coordinates to generate a visually annotated image using Python and Pillow.

What is the JSON structure required for visualizing VLM segmentation results?▼

The JSON structure required for visualizing VLM segmentation results must contain a panels array where each item provides a label and box_2d coordinates, allowing the script to read the file and map normalized coordinates onto the original image.

Does this visualization script work with Pillow for image annotation?▼

Yes, this visualization script works with Pillow as its core dependency to process image annotation, reading the JSON file and outputting the final visually annotated image for quick QA checks.

Can I use normalized coordinates from a VLM to draw boxes on original images?▼

Yes, you can use normalized coordinates from a VLM because the script reads box_2d coordinates from the JSON file and converts them into pixel coordinates to accurately draw bounding boxes on the original images.

What is the best way to create visual reports from VLM bounding box annotations?▼

The best way to create visual reports from VLM bounding box annotations is to run a Python script that overlays the box_2d data onto the source image, producing clear annotated visuals for segmentation verification in CV pipelines.