LEADERG's AI door god flags smoke; temple fire held to 10 m²

AI camera analytics flagged smoke in an empty temple office a minute before a battery fire; held to 10 m², no injuries, main hall intact.

AI system

System

AI門神:宮廟AI防火防盜系統 (AI Door God: temple AI fire-and-theft prevention system), AI image recognition over the temple's existing CCTV with alerts by SMS, LINE and email

Model

Not assessed

Release

Exact release not established.

Application developer

立達軟體科技股份有限公司 (LEADERG Inc.), Taiwan

Role

The software detected smoke in camera images and alerted temple staff.

Configuration

The image-recognition software ran on an Advantech IPC-730 computer with a SKY-QUAD-5000A-32 accelerator card. Advantech reports processing about one frame every five seconds.

Inputs and outputs

The software analyzed CCTV images and sent smoke alerts to temple staff through LINE.

Autonomy

Not assessed

Workflow

AI門神 detected smoke through cameras that already covered the temple and alerted staff through LINE. The night clerk reached the room with an extinguisher. Firefighters arrived eight minutes later.

Human oversight

A clerk attempted to extinguish the fire. Firefighters controlled it after the temple called the fire service.

Relation to AI

The system's documented contribution is the first alert. By the temple's own log it detected smoke at 05:30, a minute before the lithium batteries burst at 05:33. It messaged the front desk, the general manager and the chairman in sequence over LINE. That is how a chairman asleep in Nara came to follow the fire on the temple's cameras. The suppression was human. The clerk's extinguisher failed once the batteries burst. The Yunlin County Fire Bureau's eight trucks and eighteen firefighters, called at 05:36, controlled the fire at 06:10. UDN reports that a residential smoke alarm (住警器) sounded as well. The Third Brigade commander 陳岳謄 attributed the containment to the building's reinforced-concrete structure and the crews' quick arrival without mentioning the AI. The LINE relay from the vendor is the chairman's account, not the bureau's.

AI contribution

According to the temple’s log, AI Door God detected smoke at 05:30 and sent alerts through LINE. The chairman followed the developing fire through the temple’s cameras.

Benefits

Outcome identifiers

Not assessed

Benefit

Staff and firefighters contained the temple fire to about 10 m². No one was injured, and the main hall remained intact.

Category

Reduced Risk

Status

Not assessed

Beneficiary groups

Organizations.

Beneficiaries

Beigang Wude Temple and the people who used it.

Reach

Not assessed

Distribution and access

Not assessed

Duration

Not assessed

Expected?

Yes

Costs and harms

Not assessed

Measurement

Studies

Not assessed

Study design

Not assessed

Measurement

Not assessed

Setting and sample

Not assessed

Measures

Warning time, area damaged, injuries, and continued use of the temple.

Comparison

Assumed alternative: the temple's overnight duty staff relying on their own watch, without the AI smoke alert

Results

The temple reports that the software detected smoke about one minute before the batteries burst. A conventional smoke alarm also sounded. The AI alert’s separate effect on the damage is unmeasured.

Additional value from AI

Not assessed

Limitations

The temple’s account supplies the alert chronology. A residential smoke alarm also sounded. The fire service attributed containment to the building’s concrete structure and the firefighters’ arrival, without identifying a separate contribution from AI.

Denominators

Not assessed

Materials for replication

Not assessed

Money

Kind of value

Not assessed

Amount

Not assessed

Currency and prices

Not assessed

Period and unit

Not assessed

Realization

Not assessed

Recipients and payers

Not assessed

Gross or net

Not assessed

Included costs

Not assessed

Comparison

Not assessed

Method

Not assessed

Return and payback

Not assessed

People and places

Participants

Developer: 立達軟體科技股份有限公司 (LEADERG Inc.), Taiwan. Operator: 北港武德宮 (Beigang Wude Temple).

Jurisdictions

Benefit: Beigang, Taiwan. Developer: Taiwan. Operator: Taiwan.

Ownership and control

Not assessed

Dependencies

Not assessed

Cross-border connections

Not assessed

Geopolitical implications

Not assessed

Follow-up

Conditions for replication

Cameras need to cover rooms where a fire could start. Alerts need to reach staff who can respond within minutes. Conventional smoke alarms remain necessary.

Replication attempts

2 follow-up entries. Outcome unresolved: 2.

Later findings

25 August 2026. In August 2026, after the Zhinan Temple fire in Taipei, UDN reported Wude Temple still relying on its AI alarm system combined with the cameras, recounting the February fire as the case that proved it; no other temple adopting the system is reported.

Availability

Not assessed

Institutional decision

Not assessed

Grounds

Not assessed

Action and result

Not assessed

Success criterion

Not assessed

Evidence

Source numbers

AIBD-0155-source-0001–0012

Citations

Numbered references below.

Language

Not assessed

Source relationships

Independent, Developer

Original evidence

Not assessed

Access and versions

Not assessed

Supported findings

Not assessed

Evidence strength

Not assessed

Missing evidence

Not assessed

Grades

Assessed benefits

Organizations.

Depth

Organizations: 2.

Reasons

The fire damaged one office, while the main hall and worship areas remained usable. The temple stayed open. The fire bureau credited the concrete structure and firefighters’ arrival for containment. The AI alert provided earlier warning, but the evidence does not isolate its contribution to the damage avoided.

Observed or intended

Not assessed

Rubric

Not assessed

Assessor and date

Not assessed

Alternative assessments

Not assessed

Fill

Independently reported

Entry details

Kind

Not assessed

Field

Fire safety (religious site). Temple and cultural-asset protection: AI video analytics on existing CCTV.

Purpose

AI門神 (“AI door god: temple AI fire-and-theft prevention system”) was built to run AI image recognition over a temple's existing CCTV. It was built to detect excess smoke, flame and suspicious behavior, and to alert staff by SMS, LINE or email. CTS reported in December 2024 that Wude Temple's door-god system would sound an alarm at once on abnormal smoke. Flagging smoke before a fire was exactly its designed purpose.

Dates

Date basis unassessed: 2026 (precision unassessed). Updated: 25 October 2026.

Stage

Not assessed

Version

SHA-256: d9e50b5220d0…

References

Other identifiers

oecd_aim: 2026-02-04-4944.

Category mappings

Not assessed

Brief

Not assessed

Revisions

6 revisions. Latest: 25 October 2026.

Sources

  1. News report Independent author

    蔡維斌 / 聯合新聞網 (2026). 「財神祖廟」北港武德宮火警!AI警報響起快速撲滅…人員幸運無傷. UDN, 5 February 2026.

  2. News report Independent author

    蔡維斌 / 聯合新聞網 (2026). 強大AI!人在日本的武德宮主委被AI警報嚇醒…跨海還原火警現場. UDN, 5 February 2026.

  3. News report Independent author

    吳淑萍 / TVBS新聞網 (2026). 北港武德宮清晨竄火!鋰電池充電爆炸 「AI門神」示警救廟. TVBS, 5 February 2026.

  4. News report Independent author

    蘇榮泉 / 觀傳媒 (2026). (有影片)/AI火警系統即時示警 北港武德宮火勢10分鐘內受控. Syndicated on yam 蕃新聞, 5 February 2026.

  5. News report Independent author

    洪佳伶 / 觀傳媒 (2026). AI門神立大功!北港武德宮幹部辦公室火警 主委跨海監控、10分鐘內控制火勢. Syndicated on yam 蕃新聞, 5 February 2026.

  6. News report Independent author

    華視新聞 / CTS (2026). 「AI門神」立功! 武德宮火警偵測回報 快速控制火勢. CTS News, 5 February 2026.

  7. News report Independent author

    鄧庭婷 / NOWnews今日新聞 (2026). 北港武德宮火警!「1科技」即時通報立大功 主委:不幸中的大幸. NOWnews, 5 February 2026.

  8. News report Independent author

    莊曜聰 / 壹蘋新聞網 (2026). 北港武德宮火警損失輕微靠「兩功臣」 主委林安樂赴日收警報嘆「沒放假的命」. NextApple, 5 February 2026.

  9. News report Independent author

    蘇峯毅 / 台灣好報 (2026). 清晨火警驚魂 北港武德宮即時搶救守住信仰中心. Syndicated on yam 蕃新聞, 5 February 2026.

  10. News report Independent author

    華視新聞 / CTS (2024). AI進駐宮廟 統計人流.預防火災還能「解籤詩」. Syndicated on Yahoo奇摩新聞, 23 December 2024.

  11. News report Independent author

    聯合新聞網 UDN (2026). 指南宮大火啟示錄!雲林兩大廟分享防火經驗「要神也要人」. udn.com, 25 August 2026.

  12. Press release Developer

    Advantech (2025). 立達軟體攜手研華AI門神:創新技術助力寺廟文化資產保護. Advantech case study, 15 April 2025.