A powerful vision SDK providing image and video analysis, OCR, and spatial reasoning through Isaac vision models.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install perceptron
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install perceptron using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Perceptron Vision SDK?
Perceptron is a specialized toolkit designed for deep visual understanding, allowing AI agents to perform complex tasks such as visual Q&A, object detection, and grounded spatial reasoning. By leveraging the Isaac vision models, it provides high-accuracy analysis of images and video frames, making it an essential component for developers building sophisticated Openclaw Skills.
Unlike simple image recognition, Perceptron offers a robust Python SDK and CLI that support structured outputs, DSL composition, and in-context learning. It handles the heavy lifting of base64 conversion and coordinate normalization, allowing developers to focus on building intelligent multimodal applications.
Perceptron Vision SDK Use Cases
- Automating document data extraction with high-fidelity OCR and markdown output.
- Implementing safety monitoring systems that detect specific hazards using bounding boxes.
- Performing inventory management through automated object counting and localization.
- Enhancing accessibility by generating detailed scene captions and visual descriptions.
- Analyzing video streams by processing extracted frames for temporal insights.
How Perceptron Vision SDK Works
- Initialize the environment by installing the perceptron library and configuring the API key.
- Provide image data via local file paths or remote URLs, which the SDK automatically prepares for analysis.
- Invoke specific vision functions such as question, detect, or ocr based on the required outcome.
- Apply grounding parameters like box, point, or polygon to retrieve spatial coordinates alongside text responses.
- Process the normalized coordinate data (0-1000) or structured Pydantic models for integration into larger agentic workflows.
Perceptron Vision SDK Setup
To begin using this skill, install the package and set your environment variables:
pip install perceptron
export PERCEPTRON_API_KEY=ak_your_api_key
Perceptron Vision SDK Data Schema & Taxonomy
The skill organizes data into structured responses and spatial coordinates.
| Data Type | Description | Format |
|---|---|---|
| Text | Descriptive answers or OCR results | String / Markdown |
| Coordinates | Normalized spatial data | 0-1000 Scale |
| Boxes | Top-left and bottom-right points | [x1, y1, x2, y2] |
| Structured | Pydantic-validated responses | JSON |
Perceptron Vision SDK Advanced Features
- DSL Composition: Use the @perceive decorator to build custom multimodal workflows and system prompts.
- Grounded Spatial Reasoning: Retrieve precise pixel locations for objects using the expects parameter.
- Structured Outputs: Enforce strict response formats using Pydantic models, JSON schemas, or regex.
- Batch Processing: Efficiently analyze multiple images or video frames in a single execution loop.
- Coordinate Scaling: Built-in utilities to convert normalized model coordinates directly to image pixels.
SKILL.md
Loading
Related Openclaw Skills
Finance Watcher
A command-line tool for monitoring real-time stock and cryptocurrency prices with automated alerts and daily reporting.
Content Watcher
An AI-powered content monitoring tool that transforms RSS feeds and news sources into automated, summarized daily digests.
Epistemic Council Skill
An advanced AI governance engine for managing epistemic pipelines, validating logical claims, and auditing data integrity.
X CDP Browser Automation
A powerful browser-based automation tool for X (Twitter) that uses Chromium CDP to perform human-like social interactions without requiring official API access.
tvscreener
A powerful interface for querying real-time TradingView market data and technical indicators for global stocks and ETFs.
Feishu Knowledge Manager
An automated workflow for capturing, summarizing, and organizing web content from various platforms into a structured Feishu knowledge base.
Featured*
