How Accurate Is AI at Spotting Underground Parking Fires? Inside AIPro’s ‘FireFocus’ Field Test

Can Vision AI Secure the Golden Hour in Underground EV Fires?

With recent headlines dominated by electric vehicle (EV) fires, facility managers across residential complexes, public institutions, and commercial buildings are under immense pressure. In enclosed underground parking garages, thick toxic smoke spreads fast and fire trucks struggle to gain access. That means everything comes down to one critical factor: how quickly and accurately you can catch the very first signs of smoke and flames.

But when the stakes are this high, how reliable is AI fire detection in practice?

 

To answer that question with real-world data, vision AI specialist AIPro teamed up with EV fire response specialist S-Solution to conduct a head-to-head Fire Detection AI Model Performance Test. Moving beyond lab simulations, we tested our systems against thick smoke, live flames, and tricky visual decoys designed to trigger false alarms. Here is a look inside the field test.

---

 

Validating AI for Underground EV Parking Environments

The goal of this test was to objectively evaluate and compare AI detection models tailored for underground EV parking lots—and ultimately deliver a cost-effective, rock-solid integrated fire safety solution to the market.

• Project Overview

  • Project Name: Fire Detection AI Model Performance Comparison Test
  • Date: Friday, October 17, 2025
  • Hosted by: S-Solution Co., Ltd.
  • Participating Partner: AIPro Co., Ltd.

• Tested Solutions

  • S-Solution: AI Fire Detection Camera (ME100)
  • AIPro: AI Fire Analytics Solution (FireFocus)

  클릭하여 확대

 

From Fog Machines to Fake Campfires: 3 Real-World Test Scenarios

In the field, fire detection systems usually fail in one of two ways: missed detections when a real fire breaks out, or constant false alarms (false positives) triggered by headlights, reflections, or steam. To put both capabilities to the test, we ran three distinct scenarios:

1. Smoke Detection Accuracy (Fog Machine Test): Using a fog machine to simulate the unpredictable dispersion of early-stage smoke, we measured how quickly and accurately the AI recognized diffusing smoke plumes.
2. False Positive Evaluation (Artificial Campfire Test): We set up a prop "artificial campfire" using fabric and LED lighting to see if the AI would be fooled by fire-colored lights and flickering motion—or accurately filter them out as non-threats.
3. Live Flame & Smoke Detection (Real Brazier Test): By igniting a real brazier, we evaluated real-time detection speed and tracking stability when genuine flames and smoke occurred simultaneously.

 

Test Scenario	Core Objective & What We Measured
1. Fog Machine Test	Speed and accuracy in spotting early-stage, non-uniform smoke patterns
2. Artificial Campfire Test	False alarm immunity against fire-like lighting, colors, and movement
3. Live Brazier Fire Test	Real-time detection reliability and response speed on genuine flames and smoke

 

클릭하여 확대

 

Field Report: Voices from the AIPro Engineering Team

Engineers from AIPro’s technical team who ran the on-site testing shared their firsthand perspective on the challenges facility managers face today—and what this partnership means going forward:

"With the recent wave of EV fire incidents, facility managers at public agencies and enterprises are deeply concerned. Every time we visit trade shows or client sites, we hear the exact same thing: 'We shouldn't even allow risky EVs to park underground.' It really drives home how heavy the burden is for on-site personnel right now.

We ran this AI performance test to benchmark and validate a fire detection solution that is both cost-effective and genuinely dependable. AIPro’s core strength lies in rapid recognition and high-precision video analytics, while our partner S-Solution brings specialized hardware and response systems built specifically for EV fires.

Since the second half of 2025, our two companies have been conducting joint R&D to combine our strengths into a practical, streamlined integrated system. This field test further solidifies our partnership, and we plan to keep pushing our joint development so this combined solution can protect a wide range of industrial sites."

 

What This Means: Pairing Precision Analytics with Specialized Response

This field test goes beyond simply proving that a camera can spot a flame. It demonstrates the synergy between AIPro’s high-precision vision software (FireFocus) and S-Solution’s EV-specialized fire hardware and response tech (ME100).

Backed by ongoing joint R&D since late 2025, the two companies are accelerating the rollout of a practical, field-ready integrated fire safety system—not just for underground parking garages, but for logistics warehouses, manufacturing plants, and industrial facilities nationwide.

 

Solving Real Problems Through Field-Driven R&D

AIPro and S-Solution successfully validated their AI fire detection models across smoke, false-positive, and live-fire tests, advancing their joint R&D to deliver a practical, integrated EV fire monitoring solution.

🔎 For our team at AIPro, this test was another valuable milestone built on raw, real-world field data. Rather than building technology that only works on paper, we remain committed to hands-on, field-verified R&D—growing as an AI company that solves real problems on the ground.

---

✔ Underground parking EV fire detection ✔ AI fire and smoke detection solution ✔ AIPro FireFocus video analytics ✔ False alarm prevention in fire AI ✔ EV fire monitoring camera system ✔ Real-world AI fire performance test

---