Towards LLM-aided multimodal intelligent emergency response system

カテゴリ: 研究会(論文単位)

論文No: CMN24033

グループ名: 【C】電子・情報・システム部門 通信研究会

発行日: 2024/03/25

タイトル(英語): Towards LLM-aided multimodal intelligent emergency response system

著者名: Banerjee Anoushka(Hitachi India Limited),Mitra Abhishek(Hitachi India Limited),Ganesh Ananth(Hitachi India Limited)

著者名(英語): Anoushka Banerjee(Hitachi India Limited),Abhishek Mitra(Hitachi India Limited),Ananth Ganesh(Hitachi India Limited)

キーワード: smart city|Emergency response system|LLM-aided|multimodal|automatic alert|situational awareness|smart city|Emergency response system|LLM-aided|multimodal|automatic alert|situational awareness

要約(日本語): With the advent of numerous smart city projects, it is imperative to develop an intelligent emergency response system. The intelligent emergency response system will improve situational awareness, reduce the latency in dispensing rescue services, and overall improve citizen safety. However, at present there is a scope for substantial research endeavours towards the advancement of intelligent response systems. Apropos the exigency, we propose a large language model-aided (LLM-aided) multimodal response system. The intelligent response system primarily comprises an object recognition algorithm integrated into an LLM. We capitalize on the temporal dependency across video frames for aggregating video content to apprehend emergencies. The compounded context from video frames is passed to LLM for image-to-text information conversion.

要約(英語): With the advent of numerous smart city projects, it is imperative to develop an intelligent emergency response system. The intelligent emergency response system will improve situational awareness, reduce the latency in dispensing rescue services, and overall improve citizen safety. However, at present there is a scope for substantial research endeavours towards the advancement of intelligent response systems. Apropos the exigency, we propose a large language model-aided (LLM-aided) multimodal response system. The intelligent response system primarily comprises an object recognition algorithm integrated into an LLM. We capitalize on the temporal dependency across video frames for aggregating video content to apprehend emergencies. The compounded context from video frames is passed to LLM for image-to-text information conversion.

本誌: 2024年3月28日-2024年3月29日通信研究会

本誌掲載ページ: 89-93 p

原稿種別: 英語

PDFファイルサイズ: 1,609 Kバイト