Filipe Gama, Matěj Mísař, Lukáš Navara, Sergiu T Popescu, Matej Hoffmann
{"title":"基于视频的婴儿2D姿态自动估计:比较7种深度神经网络方法。","authors":"Filipe Gama, Matěj Mísař, Lukáš Navara, Sergiu T Popescu, Matej Hoffmann","doi":"10.3758/s13428-025-02816-x","DOIUrl":null,"url":null,"abstract":"<p><p>Automatic markerless estimation of infant posture and motion from ordinary videos carries great potential for movement studies \"in the wild\", facilitating understanding of motor development and massively increasing the chances of early diagnosis of disorders. There has been a rapid development of human pose estimation methods in computer vision, thanks to advances in deep learning and machine learning. However, these methods are trained on datasets that feature adults in different contexts. This work tests and compares seven popular methods (AlphaPose, DeepLabCut/DeeperCut, Detectron2, HRNet, MediaPipe/BlazePose, OpenPose, and ViTPose) on videos of infants in supine position and in more complex settings. Surprisingly, all methods except DeepLabCut and MediaPipe exhibit competitive performance without additional fine-tuning, with ViTPose performing the best. Next to standard performance metrics (average precision and recall), we introduce errors expressed in the neck-mid-hip (torso length) ratio and additionally study missing and redundant detections, and the reliability of the internal confidence ratings of the different methods, which are relevant for downstream tasks. Among the networks with competitive performance, only AlphaPose could run at close to real-time speed (27 fps) on our machine. We provide documented Docker containers or instructions for all the methods we used, our analysis scripts, and the processed data at https://hub.docker.com/u/humanoidsctu and https://osf.io/x465b/ .</p>","PeriodicalId":8717,"journal":{"name":"Behavior Research Methods","volume":"57 10","pages":"280"},"PeriodicalIF":3.9000,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Automatic infant 2D pose estimation from videos: Comparing seven deep neural network methods.\",\"authors\":\"Filipe Gama, Matěj Mísař, Lukáš Navara, Sergiu T Popescu, Matej Hoffmann\",\"doi\":\"10.3758/s13428-025-02816-x\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>Automatic markerless estimation of infant posture and motion from ordinary videos carries great potential for movement studies \\\"in the wild\\\", facilitating understanding of motor development and massively increasing the chances of early diagnosis of disorders. There has been a rapid development of human pose estimation methods in computer vision, thanks to advances in deep learning and machine learning. However, these methods are trained on datasets that feature adults in different contexts. This work tests and compares seven popular methods (AlphaPose, DeepLabCut/DeeperCut, Detectron2, HRNet, MediaPipe/BlazePose, OpenPose, and ViTPose) on videos of infants in supine position and in more complex settings. Surprisingly, all methods except DeepLabCut and MediaPipe exhibit competitive performance without additional fine-tuning, with ViTPose performing the best. Next to standard performance metrics (average precision and recall), we introduce errors expressed in the neck-mid-hip (torso length) ratio and additionally study missing and redundant detections, and the reliability of the internal confidence ratings of the different methods, which are relevant for downstream tasks. Among the networks with competitive performance, only AlphaPose could run at close to real-time speed (27 fps) on our machine. We provide documented Docker containers or instructions for all the methods we used, our analysis scripts, and the processed data at https://hub.docker.com/u/humanoidsctu and https://osf.io/x465b/ .</p>\",\"PeriodicalId\":8717,\"journal\":{\"name\":\"Behavior Research Methods\",\"volume\":\"57 10\",\"pages\":\"280\"},\"PeriodicalIF\":3.9000,\"publicationDate\":\"2025-09-10\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Behavior Research Methods\",\"FirstCategoryId\":\"102\",\"ListUrlMain\":\"https://doi.org/10.3758/s13428-025-02816-x\",\"RegionNum\":2,\"RegionCategory\":\"心理学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"PSYCHOLOGY, EXPERIMENTAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Behavior Research Methods","FirstCategoryId":"102","ListUrlMain":"https://doi.org/10.3758/s13428-025-02816-x","RegionNum":2,"RegionCategory":"心理学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"PSYCHOLOGY, EXPERIMENTAL","Score":null,"Total":0}
Automatic infant 2D pose estimation from videos: Comparing seven deep neural network methods.
Automatic markerless estimation of infant posture and motion from ordinary videos carries great potential for movement studies "in the wild", facilitating understanding of motor development and massively increasing the chances of early diagnosis of disorders. There has been a rapid development of human pose estimation methods in computer vision, thanks to advances in deep learning and machine learning. However, these methods are trained on datasets that feature adults in different contexts. This work tests and compares seven popular methods (AlphaPose, DeepLabCut/DeeperCut, Detectron2, HRNet, MediaPipe/BlazePose, OpenPose, and ViTPose) on videos of infants in supine position and in more complex settings. Surprisingly, all methods except DeepLabCut and MediaPipe exhibit competitive performance without additional fine-tuning, with ViTPose performing the best. Next to standard performance metrics (average precision and recall), we introduce errors expressed in the neck-mid-hip (torso length) ratio and additionally study missing and redundant detections, and the reliability of the internal confidence ratings of the different methods, which are relevant for downstream tasks. Among the networks with competitive performance, only AlphaPose could run at close to real-time speed (27 fps) on our machine. We provide documented Docker containers or instructions for all the methods we used, our analysis scripts, and the processed data at https://hub.docker.com/u/humanoidsctu and https://osf.io/x465b/ .
期刊介绍:
Behavior Research Methods publishes articles concerned with the methods, techniques, and instrumentation of research in experimental psychology. The journal focuses particularly on the use of computer technology in psychological research. An annual special issue is devoted to this field.