Testing novel facial recognition technology to identify dogs during vaccination campaigns.
Testing novel facial recognition technology to identify dogs during vaccination campaigns.
复制标题
DOI：
10.1038/s41598-023-49522-2
复制
发表时间：
2023-12-12
影响因子：
4.6
通讯作者：
中科院分区：
文献类型：
--
作者：
缺乏识别单个动物的方法可能会成为控制人畜共患病的障碍。我们为一款识别狗的手机应用程序开发了面部识别技术，并进行了现场测试，我们使用该技术来评估坦桑尼亚农村地区狂犬病的疫苗接种覆盖率。狗接种了疫苗，使用该应用程序进行了注册，并安装了微芯片。在随后的家访中，使用应用程序对狗进行登记，操作员使用该应用程序将狗分类为已接种(匹配)或未接种(未匹配)，并使用微芯片验证分类，从而确定它们的接种状态。在534只分类犬中(251只接种疫苗，283只未接种)，应用特异度为98.9%，灵敏度为76.2%，阳性预测值为98.4%，阴性预测值为82.8%。面部识别算法正确匹配了249只(99.2%)接种疫苗的狗和植入芯片的狗(真阳性)，但未能匹配两只(0.8%)接种疫苗的狗(假阴性)。操作员正确识别了186只(74.1%)已接种疫苗的狗(真阳性)和280只(98.9%)未接种疫苗的狗(真阴性)，但错误地将58只(23.1%)已接种疫苗的狗归类为不匹配(假阴性)。劣质照片和与光线相关的颜色失真导致应用感光度降低。随着开发和操作员培训，这项技术有可能成为识别狗的有用工具，并支持研究和干预计划。
A lack of methods to identify individual animals can be a barrier to zoonoses control. We developed and field-tested facial recognition technology for a mobile phone application to identify dogs, which we used to assess vaccination coverage against rabies in rural Tanzania. Dogs were vaccinated, registered using the application, and microchipped. During subsequent household visits to validate vaccination, dogs were registered using the application and their vaccination status determined by operators using the application to classify dogs as vaccinated (matched) or unvaccinated (unmatched), with microchips validating classifications. From 534 classified dogs (251 vaccinated, 283 unvaccinated), the application specificity was 98.9% and sensitivity 76.2%, with positive and negative predictive values of 98.4% and 82.8% respectively. The facial recognition algorithm correctly matched 249 (99.2%) vaccinated and microchipped dogs (true positives) and failed to match two (0.8%) vaccinated dogs (false negatives). Operators correctly identified 186 (74.1%) vaccinated dogs (true positives), and 280 (98.9%) unvaccinated dogs (true negatives), but incorrectly classified 58 (23.1%) vaccinated dogs as unmatched (false negatives). Reduced application sensitivity resulted from poor quality photos and light-associated color distortion. With development and operator training, this technology has potential to be a useful tool to identify dogs and support research and intervention programs.
登录
查看更多内容
影响因子：
3.8
作者：
Hampson K;Coudeville L;Lembo T;Sambo M;Kieffer A;Attlan M;Barrat J;Blanton JD;Briggs DJ;Cleaveland S;Costa P;Freuling CM;Hiby E;Knopf L;Leanes F;Meslin FX;Metlin A;Miranda ME;Müller T;Nel LH;Recuenco S;Rupprecht CE;Schumacher C;Taylor L;Vigilato MA;Zinsstag J;Dushoff J;Global Alliance for Rabies Control Partners for Rabies Prevention
通讯作者：
Global Alliance for Rabies Control Partners for Rabies Prevention
影响因子：
5.2
作者：
Fahrion, Anna S.;Taylor, Louise H.;Abela-Ridder, Bernadette
通讯作者：
Abela-Ridder, Bernadette
影响因子：
1.6
作者：
Crouse, David;Jacobs, Rachel L.;Tecot, Stacey R.
通讯作者：
Tecot, Stacey R.
影响因子：
9.8
作者：
Hampson K;Dushoff J;Cleaveland S;Haydon DT;Kaare M;Packer C;Dobson A
通讯作者：
Dobson A
影响因子：
3.8
作者：
Minyoo, Abel B.;Steinmetz, Melissa;Lankester, Felix
通讯作者：
Lankester, Felix