HomeSafe-Bench

Benchmark Vision-Language Models on indoor home safety hazard detection.

3.0k|463|Updated Mar 5, 2019

Search Tags:#security #ai #safety #benchmark #vlm #hazard detection

Ready to run

npx skills add https://github.com/SharpAI/DeepCamera --skill homesafe-bench

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: HomeSafe-Bench
Source: https://github.com/SharpAI/DeepCamera/tree/main/skills/analysis/homesafe-bench
Command: npx skills add https://github.com/SharpAI/DeepCamera --skill homesafe-bench

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires openai, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill automates the detection of common indoor safety hazards from camera frames, helping to identify potential risks before they cause harm.

Core Features & Use Cases

Hazard Detection: Identifies risks across 5 categories: Fire/Smoke, Electrical, Trip/Fall, Child Safety, and Falling Objects.
VLM Integration: Leverages Vision-Language Models (VLMs) for scene analysis.
Benchmarking: Provides a structured way to evaluate VLM performance on safety-critical tasks.
Use Case: A security camera system using this Skill could alert a homeowner to a potential fire hazard like a candle near curtains or an electrical risk like an overloaded power strip.

Quick Start

Run the HomeSafe-Bench skill using your VLM server at http://localhost:5405.

Frequently Asked Questions about HomeSafe-Bench

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

FAQPage Schema

How do I evaluate Vision-Language Models for indoor safety hazard detection?▼

You can evaluate Vision-Language Models for indoor safety hazard detection by benchmarking their performance on static camera frames across fire, electrical, trip/fall, child safety, and falling object categories using AI-generated fixture images.

What types of home safety hazards can VLMs identify from camera frames?▼

VLMs can identify home safety hazards across five categories: Fire/Smoke, Electrical, Trip/Fall, Child Safety, and Falling Objects from static camera frames to help prevent potential risks.

Do I need a Node.js environment to run VLM safety benchmarks?▼

Yes, you need a Node.js environment and a running VLM server to execute the safety hazard detection benchmark and evaluate model performance on camera frame analysis.

Can I use this benchmark to test VLMs on fire and electrical hazard detection?▼

Yes, you can test VLMs on fire and electrical hazard detection by running the benchmark against your VLM server to evaluate scene analysis accuracy for safety-critical tasks.

What's the best way to benchmark VLM performance on child safety hazard detection?▼

The best way to benchmark VLM performance on child safety hazard detection is to evaluate models against AI-generated fixture images that simulate indoor hazard scenarios across multiple safety categories.

Does the hazard detection benchmark support real-time video stream analysis?▼

No, the hazard detection benchmark evaluates VLMs using static camera frames rather than real-time video streams, providing structured performance metrics for safety-critical scene analysis.