https://lear.inrialpes.fr/local/wiki/ 2023-09-06T10:25:58+02:00 https://lear.inrialpes.fr/local/wiki/ https://lear.inrialpes.fr/local/wiki/lib/tpl/dokuwiki/images/favicon.ico text/html 2016-02-26T17:08:04+02:00 https://lear.inrialpes.fr/local/wiki/doku.php?id=vidscreener:devlog&rev=1456502884&do=diff Vidscreener Todo Dataset statistics * once this dataset exists, create a function to draw a webpage with statistics: * matrix with number of videos per suggested class + presence of other classes * number of keysframes/temporal instances etc (total + per suggested class) text/html 2016-01-16T10:02:49+02:00 https://lear.inrialpes.fr/local/wiki/doku.php?id=vidscreener:guidelines&rev=1452934969&do=diff ====== Spatial Annotation Tutorial ====== Welcome to the spatial annotation tutorial. A set of keyframes where a human performs a specific action has been automitacally selected. The goal is to annotate for each one: * a bounding box around the actor, * a bounding box around the object (if any), * the pose of the upper body of the actor. Here is an overview of the interface. {{:vidscreener:annotation.png?2000|}} ==== Bounding Box ==== The first task consists in drawing a bounding box … text/html 2015-11-12T17:37:30+02:00 https://lear.inrialpes.fr/local/wiki/doku.php?id=vidscreener:main&rev=1447346250&do=diff ====== Vidscreener ====== Collaborative video annotation tool, written as a Python web server with a Javascript/HTML5 front-end. The action annotation task is separated in 3 tasks: * class annotation: to ensure a video contains the suggested action, also an opportunity to find other classes * temporal annotation: videos tagged as positive reach this point * (**todo**) spatial annotation: relying on the temporal annotation, annotate actors/objects/body joints as bounding boxes, points...…