Page 31 - OPCFHK Annual Report 2020-21
P. 31

借助人工智能技術保護蘇眉                            Utilising the Power of Artificial Intelligence
            二零二一年三月,保育基金開始資助一項有                     In March 2021, OPCFHK began funding a computer-aided facial recognition
            關臉部辨識系統的研究,透過人工智能技術                     project that employed artifi cial intelligence (AI) technology to combat the illegal
            識別不同蘇眉的臉孔,讓用家得以辨別本地                     trade of humphead wrasse. As a novel solution to improve enforcement of the
            市場上出售的蘇眉是否經由合法途徑進口,                     Convention on International Trade in Endangered Species of Wild Fauna and
            以加強執行《瀕危野生動植物種國際貿易公                     Flora (CITES), computer-aided facial recognition can help users distinguish
            約》(CITES),打擊非法蘇眉貿易。                     between legally and illegally imported humphead wrasse at Hong Kong's local
                                                    seafood markets.

            化創意理論為實質行動                              Putting the Theory into Practice
            研究團隊研發出能成功辨認不同蘇眉的臉部                     The research team developed identification models that could match
            特色的識別系統,準確率達百分之九十六點                     individual humphead wrasse based on their facial markings, from which a
            四。有關應用程式的原型已經提交漁農自然                     trained model with a high accuracy of 96.4% was developed. The team then
            護理署(漁護署),公眾亦可透過 Google                  developed a prototype app for the Agriculture, Fisheries and Conservation
            Play 商店及蘋果 App Store 下載。用家只要            Department (AFCD) and the general public, which is currently available on the
            輸入商店資料及上載蘇眉照片,應用程式即                     Google Play Store and Apple App Store. This app allows users to input shop
            會自動比對數據庫資料,查找相應的記錄,                     information and humphead wrasse photos, then matches the most likely
                                                    humphead wrasse from the database. This empowers the pubic to confi rm
            讓公眾得以確認所購買的蘇眉是否經由合法                     that the fi sh they are buying are from legal sources, and to report suspicious
            途徑進口,並透過應用程式舉報可疑個案。                     sales via the app. Every new photo added to the app will assist the team and
            每一張上載相片都有助研究團隊及漁護署構                     AFCD in building a robust database. In the future, both the database and the
            建數據庫,而用家的意見,亦有助優化數據                     app will continue to undergo optimisation based on user feedback.
            庫和應用程式。



              學名                Cheilinus undulatus
              Scienti c Name

              概要                ί࠰ಥਯርᘽ޵̀඲͡ჯ஢̙ᗇdϾ˲̥ঐܲ஢̙ᗇܸ׼ࠢᕘ၍Ϟي၇dʔཀ̹ࠦ̈ਯٙᘽ޵ᅰඎۍʔ̥׵Ϥdԑᗇڢج൱׸ਪ
              Overview          ᕚπί
                                Hong Kong only allows limited, licensed sales of humphead wrasse, but sales discrepancies indicate that illegal trade most
                                likely is occurring
              保育狀況              ᖏΚ
              Conservation Status  Endangered*
              估計數量              ܵᚃಯˇ
              Estimated Population  Decreasing
              棲息地類型             ૵ऎ੭
              Type of Habitat   Marine neritic
              主要威脅              ڢج࠮ي൱׸
              Major Threats     Illegal food trade

             * 根據世界自然保護聯盟瀕危物種紅色名錄
               According to the IUCN Red List of Threatened Species

            你也可以出一分力!
            You Can Help!

                         立即掃描二維碼,下載應用程式,識別蘇眉是否合法進口,一起以
                         行動對抗非法貿易,守護瀕危物種。
                         Scan this code to download the app now, and use it to verify the
                         source of the humphead wrasse you consume. Let's work together
                         to combat the illegal trade of threatened species.

                                                                                       IOS            Android










                                                                                                                 29
   26   27   28   29   30   31   32   33   34   35   36