skills.rest
bensonlee5
bensonlee5COMMUNITY
Explore all skills by this author →
GitHub

qwen35-4b-nf4

Answer natural language queries about RGB scenes and video frames.

Updated Jul 5, 2026
Search Tags:#multimodal#robotics#scene understanding#scene query#video language model
One-click install
npx skills add https://github.com/bensonlee5/openral --skill qwen35-4b-nf4
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: qwen35-4b-nf4
Source: https://github.com/bensonlee5/openral/tree/main/rskills/qwen35-4b-nf4
Command: npx skills add https://github.com/bensonlee5/openral --skill qwen35-4b-nf4

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill allows you to query and understand scenes captured by RGB cameras, providing detailed textual descriptions and answering specific questions about the scene.

Core Features & Use Cases

  • Scene Query: Ask about objects, spatial relations, and task progress in a scene.
  • Video Support: Process and understand video sequences, extracting insights over time.
  • Embodiment-Agnostic: Works with any robot or embodiment with an RGB camera.
  • Use Case: Use this Skill to determine if a robot's task is complete by asking if it has grasped an object or if a task is progressing as expected.

Quick Start

Install the skill and query the scene: ral skill install hf://OpenRAL/rskill-qwen35-4b-nf4 and then query_scene -q "Has the robot grasped the mug?"

Frequently Asked Questions about qwen35-4b-nf4

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

FAQPage Schema
How do I use a video language model for scene understanding and querying?▼

Scene understanding with a video language model involves passing RGB images or video frames and natural language queries to extract grounded textual descriptions and answers about objects and spatial relations.

Can I check robot task progress by asking questions about an RGB camera stream?▼

Yes, you can query an RGB camera stream to determine if a robot has grasped an object or if a task is progressing as expected by processing the frames with a multimodal video language model.

What is the best way to extract spatial relations and task progress from video frames?▼

The best way to extract spatial relations and task progress from video frames is by applying a natively-multimodal video language model that accepts natural language queries and returns grounded textual responses.

Does this scene query Skill work with any robot embodiment?▼

Yes, the scene query Skill is embodiment-agnostic, meaning it works with any robot or embodiment equipped with an RGB camera stream and a natural language processing backend.

How do I install and run a scene query for visual inspection workflows?▼

To run a scene query for visual inspection workflows, install the Skill using `ral skill install hf://OpenRAL/rskill-qwen35-4b-nf4` and execute `query_scene -q "Your question"`.

Do I need an RGB camera stream to use multimodal scene understanding tools?▼

Yes, an RGB camera stream is required as input to provide the visual data necessary for the multimodal video language model to process natural language queries and return textual responses.

Related Skills

bensonlee5bensonlee5

robometer-4b

Monitor robotic simulation task progress and success from RGB video frames.

Community
Advanced
kandengkandeng
2

Aiya Robot Controller

Convert natural language commands into Aiya humanoid robot movement, gesture, and speech actions.

Community
Intermediate
lyingparachutelyingparachute

to-tickets

Convert project plans and conversations into structured tickets with dependencies.

Community
Intermediate
Embassy-of-the-Free-MindEmbassy-of-the-Free-Mind
14

extract-images

Extract illustrations from historical book scans and generate structured metadata.

Official
Advanced