MTR Corporation Limited / The Hong Kong University of Science and Technology 港鐵公司及香港科技大學
Product得獎產品
Public transport ridership simulation – MTR to Keep Cities Moving The Unprecedented Strategy and Intelligence in Railway Planning 「鐵路營運客流量預測及規劃 以港鐵推動城市前行」
Description產品簡介
MTR and HKUST applied a big data calibrated digital twin to plan a resilient metro public transit network, adaptive to service disruptions and crowding conditions. Our advanced large-scale dynamic simulation model is a digital twin of Hong Kong’s entire public transportation system, this agent-based digital twin is currently the largest model of its type to have ever developed for Hong Kong. It can simulate the daily trips of over 4 million active travellers within the city throughout the day. The model was calibrated with an extensive dataset, it used over 6 million time-precise daily station entry-exit trips of more than 8000 station pairs within the MTR metro network. This method of forecasting ridership affords MTR déjà vu intelligence, it was applied to a planned service disruption; it generated projections that are precise to the second, which have never been available before, and enabled MTR operators to actively devise action plans ahead of time, better allocate finite resources and staffing. As such, MTR can ensure a safe operational environment, conduct future planning intelligently and systematically to keep cities moving. 香港科技大學－港鐵聯合實驗室的項目團隊，應用大數據校準的數位孿生來規劃地鐵網絡，以應對服務中斷和擁擠狀況。 該先進的大規模動態模擬模型是香港整個公共交通系統的數位雙胞胎，這個基於代理的數位雙胞胎，目前是香港有史以來開發的最大型模型。它可以模擬超過 400萬名活躍旅客在城市內全天的日常出行。該模型是基於一個廣泛的數據集進行校準的，使用了超過600萬個精確時間的每日車站進出行程，涵蓋了超過8,000對車站之間的港鐵地鐵網絡。 校準後的數位孿生模型被用來模擬港鐵觀塘線計劃於 2024 年 7 月 28 日部分關閉的影響。該模型預測每秒鐘的列車負載、月台負載和轉運模式，分析各轉乘站的人流分布及擁擠情況。團隊發現太子站在高峰會聚集大量乘客，因此港鐵採取對應積極措施鼓勵乘客分流至其他轉乘站。 在數位孿生中可以模擬預設場景，從新地鐵線的開通、至車站或線路服務暫停。模型中如此精確的流量預測是前所未有，港鐵能夠從此作適當規劃，調度資源如列車頻率和車站人員來滿足運營的需要，並推動城市前行。
APICTA 2024 Winner 大獎得主
Category類別
Consumer – Digital Marketing / Advertising 消費者類別－數碼市場推廣／廣告
Try.Eat! is an incredibly easy-to-use influencer matching platform, empowering brands and merchants to seamlessly collaborate with verified influencers through a SaaS service. On Try.Eat!, merchants can efficiently connect with suitable influencers, achieve marketing goals effortlessly, and enjoy flexible pricing models, starting with a free option. Leveraging advanced technology, Try.Eat! simplifies influencer marketing, helping brands and merchants attain significant marketing success. Since its launch in 2021, Try.Eat! has rapidly grown, featuring over 8,000 verified influencers and KOLs across major social media platforms such as Instagram, Facebook, YouTube, OpenRice, Dianping, and Xiaohongshu. Currently trusted by over 500 F&B merchants and brands, notable clients include Grand Emperor Hotel, UNY Hong Kong, APITA, Jurlique, SuperFood Lab, Green Common, Ruby Tuesday, Outback Steak House, Brick Lane, HABITŪ, A Nice Gift, and Mammy Pancake. Try.Eat！是一個極易使用的網紅行銷配對平台，讓客戶能輕鬆地與經過驗證的網紅合作。Try.Eat！提供靈活的收費模式並從免費開始，幫助商戶快速找到合適的網紅，更輕鬆達到行銷目標。自2021年推出以來，Try.Eat！迅速成長，擁有超過8,000名經過驗證的網紅，覆蓋Instagram、Facebook、YouTube、OpenRice、大眾點評和小紅書等主要平台，並受到超過500家餐飲商戶與品牌的信賴，包括英皇娛樂酒店、UNY Hong Kong、APITA、Jurlique、SuperFood Lab、吃茶三千、Green Common、Rube Tuesday、Outback Steak House、Brick Lane、HABITŪ、賞茶和媽咪雞蛋仔等。
APICTA 2024 Winner 大獎得主
Category類別
Consumer – Tourism & Hospitality 消費者類別－旅遊與款待
Organisation得獎者
Maphive Technology Limited 蜂圖科技有限公司
Product得獎產品
CityGeni – Smart City App Specializing in Barrier-Free Navigation CityGeni「智慧城市應用程式，為旅客和市民提供無障礙導航」
Description產品簡介
Mapxus is a leading GIS Intelligence Platform and PropTech Solutions Provider specializing in indoor mapping and navigation. Focused on creating barrier-free, inclusive urban environments, Mapxus uses advanced technology to deliver seamless indoor-outdoor connectivity, helping businesses, individuals, and city planners navigate complex spaces efficiently. With a strong focus on accessibility, Mapxus offers adaptable solutions that serve users of all abilities. Its software-based, sustainable approach to indoor mapping is cost-effective and easy to implement, breaking traditional barriers to entry. Mapxus equips businesses and communities to unlock their full potential using existing indoor location data, transforming spaces and enriching urban living experiences worldwide. Mapxus 是領先的 GIS 智慧平台與房地產科技解決方案供應商，專注於室內地圖與導航技術。透過無縫連接，Mapxus 協助企業及城市規劃師導航空間，推動無障礙環境。其解決方案具成本效益，幫助善用室內定位數據，提升城市生活質量。
APICTA 2024 Winner 大獎得主
Category類別
Inclusions & Community Services – Sustainability and Environment 社區共融及社會服務類別－可持續性與環境
Organisation得獎者
Green AI Technology Limited 綠色人工智能科技有限公司
Product得獎產品
AI-Driven Sorting System 「基於人工智能的分揀系統」
Description產品簡介
Green AI, dedicates its effort to deliver the first AI-driven sorting system that holds over 500k waste data images in its own database, and covering more than 40 waste categories in Hong Kong. The AI-driven sorting system is compatible with various types of sorting equipment, including optical sorters and robotic arms. Two key features distinguish Green AI’s system from existing high-tech recognition units on the market. First, it integrates multi-sensors into its AI vision unit, allowing it to perceive waste similarly to human eyes. This capability enables it to recognize not only material information but also attributes such as colour, brand, size, food and non-food grades, and functional use. Second, the system’s precise recognition abilities are significantly enhanced by its AI self-learning capability. The AI sorting algorithm developed can automatically identify localized features of items using deep filters and incorporates part attention mechanisms (that determines the relative importance of each part in a sequence relative to the other parts) This allows the AI to focus on specific parts of objects, thereby streamlining data collection, training, and learning times. As a result, these two key features of the AI-driven sorting system are well-equipped to address different types of waste management challenges and pave the way for improved recycling practices. Green AI 致力於提供首個人工智慧分類系統，該系統的自有資料庫中擁有超過 50萬張廢物資料圖像，涵蓋香港40多個廢物類別。 人工智慧分類系統可與各種類型的分類設備相容，包括光學分選機和機械臂。Green AI的系統與市場上現有的高科技識別裝置有兩個關鍵特點。首先，它將多感測器集成到其人工智慧視覺單元中，使其能夠像人眼一樣感知廢物。此功能使其不僅能夠識別材料資訊，還能識別顏色、品牌、尺寸、食品和非食品等級以及功能用途等屬性。 其次，該系統的精確識別能力因其人工智慧自學能力而得到顯著增強。開發的人工智慧分類演算法可以使用深度篩檢程式自動識別物品的局部特徵，並結合部分注意機制（確定序列中每個部分相對於其他部分的相對重要性），這使人工智慧能夠專注於物體的特定部分，從而簡化資料收集、訓練和學習時間。 因此，人工智慧分類系統的這兩個關鍵功能能夠很好地解決不同類型的廢物管理挑戰，並為改善回收實踐鋪平道路
APICTA 2024 Winner 大獎得主
Category類別
Public Sector and Government – Government & Citizen Services 公營機構及政府類別－政府及市民服務
Organisation得獎者
Transport Department, The Government of the HKSAR / Autotoll Limited / Ove Arup & Partners Hong Kong Limited 香港特別行政區政府運輸署、快易通有限公司及奧雅納工程顧問
Product得獎產品
HKeToll 「不停車繳費系統『易通行』」
Description產品簡介
Hong Kong’s first free-flow tolling system – HKeToll – is a key smart mobility initiative promulgated in the Hong Kong Smart City Blueprint, aimed at utilising technologies to enhance the overall efficiency of the city’s transport system. It replaced the traditional manual and automatic toll booths, allowing motorists to pay tunnel tolls without stopping at toll booths. Construction of new roadside structure is not required, reducing capital cost and time for implementation. The service mainly consists of HKeToll system and Toll Tag. Utilising Radio Frequency Identification (RFID) technology, combined with Automatic Number Plate Recognition and LiDAR, the system reads the unpowered Toll Tags affixed to vehicles’ windscreens, and tolls are automatically collected from the pre-set payment accounts. Additionally, the back-end system running in the Government Cloud is flexible, allowing for rapid addition of new functions to enhance service efficiency. Motorists can manage their accounts and settle tolls conveniently through HKeToll website or mobile app, using various electronic payment methods. By seamlessly integrating advanced technologies and innovative solutions, HKeToll provides a brand new experience of smart mobility for all motorists in Hong Kong. 「易通行」不停車繳費系統是《香港智慧城市藍圖》其中一項智慧出行措施，旨在運用科技提高運輸系統的整體效率。此系統全面取代傳統的人手及自動收費亭，讓車輛無須在收費亭停車便可繳費，並無需新建道路建築物。 它主要由「易通行收費系統」及「繳費貼」兩部分組成。透過無線射頻識別技術，配合自動車牌識別技術和光學雷達，系統可讀取貼於車輛擋風玻璃上無須接駁電源的繳費貼，便可從預設的繳費戶口中自動扣除隧道費。同時，在政府雲端上的「易通行」後台系統具靈活性，能視乎情況快速增設新功能，進一步提升服務效率。 駕駛者可透過「易通行」網站或流動應用程序，選擇使用多種電子支付方式，方便管理帳戶及付費。透過無縫整合先進技術和創新解決方案，「易通行」為香港所有駕駛者提供全新的智慧移動體驗。
APICTA 2024 Winner 大獎得主
Category類別
Senior Student: Application 學生類別－高中學生：應用
Organisation得獎者
ELCHK Lutheran Secondary School 基督教香港信義會信義中學
Product得獎產品
A-EYE AI visual impaired shopping assistant 「人工智能視障人士購物助理」
Description產品簡介
A-EYE is a revolutionary assistive technology designed to enhance the convenience and independence of visually impaired individuals during their everyday shopping experiences. By integrating advanced technologies such as Python, YOLOv8, PyTorch, OpenCV, and Raspberry Pi, A-EYE offers a comprehensive range of functionalities, including product recognition, price comparison, discount inquiry, currency recognition, navigation guidance, and safety assurance, enabling them to confidently take control of their shopping process. The A-EYE product series includes three different versions to meet various usage scenarios and needs In addition to providing convenience for visually impaired users, A-EYE can also bring positive value to partnering merchants. By integrating with the A-EYE system, merchants can better understand the needs of visually impaired consumers, offer targeted services and discounts, and gain valuable user insights and shopping data to support business development. We welcome more merchants to join the A-EYE partnership. 「A-EYE人工智能視障人士購物助理」是一項革命性的輔助技術，旨在增強視障人士在日常購物體驗中的便利性和獨立性。通過整合Python、YOLOv8、PyTorch、OpenCV和Raspberry Pi等先進技術，A-EYE提供了一系列全面的功能，包括產品識別、價格比較、折扣查詢、貨幣識別、導航指導和安全保障，使他們能夠自信地掌控購物過程。 A-EYE產品系列包括三個不同版本，以滿足各種使用場景和需求。 除了為視障用戶提供便利外，A-EYE還可以為合作商家帶來積極的價值。通過與A-EYE系統的整合，商家可以更好地瞭解視障消費者的需求，提供針對性的服務和折扣，並獲得有價值的使用者資料及購物數據，以支援業務發展。我們歡迎更多商家加盟成為A-EYE的合作夥伴。
APICTA 2024 Winner 大獎得主
Category類別
Tertiary Student: Application 學生類別－專上學生項目：應用
Organisation得獎者
Hong Kong Baptist University 香港浸會大學
Product得獎產品
From Romanised Nepali to Pure Nepali: AI Transliteration Using Transformer Architecture for Low-Resourced Languages 「從羅馬化尼泊爾語到純尼泊爾語：使用Transformer架構為進行資源匱乏語言的AI音譯」
Description產品簡介
This innovative transliteration system bridges the gap between Romanized Nepali and the Devanagari script, addressing the challenges faced by low-resource languages like Nepali. With over 35 million speakers, Nepali’s digital presence is limited, primarily due to the predominance of Romanized text in informal communications. The system employs advanced tools and techniques, utilising Python, TensorFlow, and PyTorch, along with transformer-based architectures like mT5, mBART, Llama 3 and GPT 3.5 Turbo. These technologies facilitate high-accuracy, context-aware transliteration while overcoming the limitations of traditional rule-based systems. By leveraging attention mechanisms, the system adapts to the fluid nature of informal language, ensuring accurate transliterations even for variations like “babu” and “baabu.” Key achievements include the creation of the world’s largest dataset of pure Nepali and Romanized Nepali sentence pairs and the development of a novel Nepali sentence-level deduplication technique. Additionally, the integration of Parameter-Efficient Fine-Tuning (PEFT) and Low-Rank Adaptation (LoRA) techniques ensures computational efficiency while maintaining high performance. This project not only enhances the digital accessibility of Nepali speakers but also sets a precedent for expanding similar solutions to other low-resource languages globally. 這種創新的音譯系統彌合了羅馬化尼泊爾語和梵文文字之間的差距，解決了尼泊爾語等資源匱乏語言所面臨的挑戰。尼泊爾語的使用者超過3,500萬，其數位影響力有限，這主要是由於羅馬化文本在非正式交流中佔主導地位。該系統採用先進的工具和技術，利用Python、TensorFlow和PyTorch，以及基於Transformer神經網絡的架構，如mT5、mBART、Llama 3和GPT 3.5 Turbo。這些技術促進了高精度、上下文感知的音譯，同時克服了傳統基於規則的系統的限制。 透過利用注意力機制，該系統適應非正式語言的流動性，即使是「babu」和「baabu」等變體也能確保準確的音譯。主要成就包括創建了世界上最大的純尼泊爾語和羅馬化尼泊爾語句子對資料集，以及開發了一種新穎的尼泊爾語句子級重複資料刪除技術。此外，參數高效微調（PEFT）和低秩適應（LoRA）技術的整合確保了運算效率，同時保持高效能。該計畫不僅增強了尼泊爾語使用者的數位可近性，還開創了將類似解決方案擴展到全球其他資源匱乏語言的先例。
Hong Kong Institute of Information Technology at IVE (Tuen Mun) 香港資訊科技學院（屯門）
Product得獎產品
SAFERIN 「智護老」
Description產品簡介
Elderly care homes face severe staff shortages, with each caregiver handling up to 20 seniors. SAFERIN for Elderly integrates three key functions into a one-stop automated solution that reduces caregiver workload, improves service quality, and ensures proper care for the elderly. SAFERIN can reduce seniors’ waiting times by 60%, enabling faster response to their needs. The saved time allows caregivers to assist more seniors, improving productivity. Concurrently, SAFERIN provides the manpower equivalent of a 75% headcount increase while saving 55% in labor costs, effectively alleviating understaffing. SAFERIN functions include proactively analyzing seniors’ needs with voice and AI to automatically prioritize tasks; detecting falls and anomalies via sensors; monitoring health data through IoT and assessing risks via machine learning for clinicians. 智護老為安老院舍提供一站式自動化解決方案，以減輕護理員負擔，提高服務素質，確保長者得到適當照顧。智護老能把長者等候時間減少六成，需求得以更快回應；節省時間讓護理員可協助更多長者，提升工作效率。與此同時，智護老相當於為院舍額外增聘了75%人手，節省約55%人力成本，有效緩解人手短缺。 智護老運用語音及人工智能，主動分析長者需求並整理優先次序；利用傳感器檢測長者跌倒等異常；通過物聯網監察長者健康數據，並利用機器學習評估風險，供醫護人員參考。
